Authors: Juliana Resplande Sant'anna Gomes, Arlindo Rodrigues Galv\~ao Filho
Abstract: The accelerated dissemination of disinformation often outpaces the capacity for manual fact-checking, highlighting the urgent need for Semi-Automated Fact-Checking (SAFC) systems. Within the Portuguese language context, there is a noted scarcity of publicly available datasets that integrate external evidence, an essential component for developing robust AFC systems, as many existing resources focus solely on classification based on intrinsic text features. This dissertation addresses this gap by developing, applying, and analyzing a methodology to enrich Portuguese news corpora (Fake.Br, COVID19.BR, MuMiN-PT) with external evidence. The approach simulates a user's verification process, employing Large Language Models (LLMs, specifically Gemini 1.5 Flash) to extract the main claim from texts and search engine APIs (Google Search API, Google FactCheck Claims Search API) to retrieve relevant external documents (evidence). Additionally, a data validation and preprocessing framework, including near-duplicate detection, is introduced to enhance the quality of the base corpora.
Authors: Yao Ge, Sudeshna Das, Yuting Guo, Abeed Sarker
Abstract: Biomedical named entity recognition (NER) is a high-utility natural language processing (NLP) task, and large language models (LLMs) show promise particularly in few-shot settings (i.e., limited training data). In this article, we address the performance challenges of LLMs for few-shot biomedical NER by investigating a dynamic prompting strategy involving retrieval-augmented generation (RAG). In our approach, the annotated in-context learning examples are selected based on their similarities with the input texts, and the prompt is dynamically updated for each instance during inference. We implemented and optimized static and dynamic prompt engineering techniques and evaluated them on five biomedical NER datasets. Static prompting with structured components increased average F1-scores by 12% for GPT-4, and 11% for GPT-3.5 and LLaMA 3-70B, relative to basic static prompting. Dynamic prompting further improved performance, with TF-IDF and SBERT retrieval methods yielding the best results, improving average F1-scores by 7.3% and 5.6% in 5-shot and 10-shot settings, respectively. These findings highlight the utility of contextually adaptive prompts via RAG for biomedical NER.
Authors: Lei Jiang, Fan Chen
Abstract: Neural scaling laws have driven the development of increasingly large language models (LLMs) by linking accuracy improvements to growth in parameter count, dataset size, and compute. However, these laws overlook the carbon emissions that scale exponentially with LLM size. This paper presents \textit{CarbonScaling}, an analytical framework that extends neural scaling laws to incorporate both operational and embodied carbon in LLM training. By integrating models for neural scaling, GPU hardware evolution, parallelism optimization, and carbon estimation, \textit{CarbonScaling} quantitatively connects model accuracy to carbon footprint. Results show that while a power-law relationship between accuracy and carbon holds, real-world inefficiencies significantly increase the scaling factor. Hardware technology scaling reduces carbon emissions for small to mid-sized models, but offers diminishing returns for extremely large LLMs due to communication overhead and underutilized GPUs. Training optimizations-especially aggressive critical batch size scaling-help alleviate this inefficiency. \textit{CarbonScaling} offers key insights for training more sustainable and carbon-efficient LLMs.
Authors: Aamod Thakur, Ajay Nagpal, Atharva Savarkar, Kundeshwar Pundalik, Siddhesh Dosi, Piyush Sawarkar, Viraj Thakur, Rohit Saluja, Maunendra Sankar Desarkar, Ganesh Ramakrishnan
Abstract: While model architecture and training objectives are well-studied, tokenization, particularly in multilingual contexts, remains a relatively neglected aspect of Large Language Model (LLM) development. Existing tokenizers often exhibit high token-to-word ratios, inefficient use of context length, and slower inference. We present a systematic study that links vocabulary size, pre-tokenization rules, and training-corpus composition to both token-to-word efficiency and model quality. To ground our analysis in a linguistically diverse context, we conduct extensive experiments on Indic scripts, which present unique challenges due to their high script diversity and orthographic complexity. Drawing on the insights from these analyses, we propose a novel algorithm for data composition that balances multilingual data for tokenizer training. Our observations on pretokenization strategies significantly improve model performance, and our data composition algorithm reduces the average token-to-word ratio by approximately 6% with respect to the conventional data randomization approach. Our tokenizer achieves more than 40% improvement on average token-to-word ratio against stateof-the-art multilingual Indic models. This improvement yields measurable gains in both model performance and inference speed. This highlights tokenization alongside architecture and training objectives as a critical lever for building efficient, scalable multilingual LLMs
Authors: Zhanye Luo, Yuefeng Han, Xiufan Yu
Abstract: Large language models (LLMs) generate text embeddings from text data, producing vector representations that capture the semantic meaning and contextual relationships of words. However, the high dimensionality of these embeddings often impedes efficiency and drives up computational cost in downstream tasks. To address this, we propose AutoEncoder-Augmented Learning with Text (AEALT), a supervised, factor-augmented framework that incorporates dimension reduction directly into pre-trained LLM workflows. First, we extract embeddings from text documents; next, we pass them through a supervised augmented autoencoder to learn low-dimensional, task-relevant latent factors. By modeling the nonlinear structure of complex embeddings, AEALT outperforms conventional deep-learning approaches that rely on raw embeddings. We validate its broad applicability with extensive experiments on classification, anomaly detection, and prediction tasks using multiple real-world public datasets. Numerical results demonstrate that AEALT yields substantial gains over both vanilla embeddings and several standard dimension reduction methods.
Authors: Ying Liu, Can Li, Ting Zhang, Mei Wang, Qiannan Zhu, Jian Li, Hua Huang
Abstract: The conversational capabilities of large language models hold significant promise for enabling scalable and interactive tutoring. While prior research has primarily examined their capacity for Socratic questioning, it often overlooks a critical dimension: adaptively guiding learners based on their cognitive states. This study shifts focus from mere question generation to the broader instructional guidance capability. We ask: Can LLMs emulate expert tutors who dynamically adjust strategies in response to learners' understanding? To investigate this, we propose GuideEval, a benchmark grounded in authentic educational dialogues that evaluates pedagogical guidance through a three-phase behavioral framework: (1) Perception, inferring learner states; (2) Orchestration, adapting instructional strategies; and (3) Elicitation, stimulating proper reflections. Empirical findings reveal that existing LLMs frequently fail to provide effective adaptive scaffolding when learners exhibit confusion or require redirection. Furthermore, we introduce a behavior-guided finetuning strategy that leverages behavior-prompted instructional dialogues, significantly enhancing guidance performance. By shifting the focus from isolated content evaluation to learner-centered interaction, our work advocates a more dialogic paradigm for evaluating Socratic LLMs.
Authors: Xiaoyuan Zhu, Muru Zhang, Ollie Liu, Robin Jia, Willie Neiswanger
Abstract: Modern large language models often encode sensitive, harmful, or copyrighted knowledge, raising the need for post-hoc unlearning-the ability to remove specific domains of knowledge from a model without full retraining. A major bottleneck in current unlearning pipelines is constructing effective forget sets-datasets that approximate the target domain and guide the model to forget it. In this work, we introduce a scalable, automated approach to generate high-quality forget sets using language models themselves. Our method synthesizes textbook-style data through a structured prompting pipeline, requiring only a domain name as input. Through experiments on unlearning biosecurity, cybersecurity, and Harry Potter novels, we show that our synthetic datasets consistently outperform the baseline synthetic alternatives and are comparable to the expert-curated ones. Additionally, ablation studies reveal that the multi-step generation pipeline significantly boosts data diversity, which in turn improves unlearning utility. Overall, our findings suggest that synthetic datasets offer a promising path toward practical, scalable unlearning for a wide range of emerging domains without the need for manual intervention. We release our code and dataset at https://github.com/xyzhu123/Synthetic_Textbook.
Authors: Zijian Chen, Xueguang Ma, Shengyao Zhuang, Ping Nie, Kai Zou, Andrew Liu, Joshua Green, Kshama Patel, Ruoxi Meng, Mingyi Su, Sahel Sharifymoghaddam, Yanxi Li, Haoran Hong, Xinyu Shi, Xuye Liu, Nandan Thakur, Crystina Zhang, Luyu Gao, Wenhu Chen, Jimmy Lin
Abstract: Deep-Research agents, which integrate large language models (LLMs) with search tools, have shown success in improving the effectiveness of handling complex queries that require iterative search planning and reasoning over search results. Evaluations on current benchmarks like BrowseComp relies on black-box live web search APIs, have notable limitations in (1) fairness: dynamic and opaque web APIs hinder fair comparisons and reproducibility of deep research methods; (2) transparency: lack of control over the document corpus makes it difficult to isolate retriever contributions. In other words, the current evaluations may compare a complete deep research system at a given time, but they do not foster well-controlled experiments to provide insights into the capability of underlying deep research LLMs. To address these challenges, we introduce BrowseComp-Plus, a benchmark derived from BrowseComp, employing a fixed, carefully curated corpus. Each query in BrowseComp-Plus includes human-verified supporting documents and mined challenging negatives, enabling controlled experimentation. The benchmark is shown to be effective in distinguishing the performance of deep research systems. For instance, the open-source model Search-R1, when paired with the BM25 retriever, achieves 3.86% accuracy, whereas the GPT-5 achieves 55.9%. Integrating the GPT-5 with the Qwen3-Embedding-8B retriever further enhances its accuracy to 70.1% with fewer search calls. This benchmark allows comprehensive evaluation and disentangled analysis of deep research agents and retrieval methods, fostering insights into retrieval effectiveness, citation accuracy, and context engineering in Deep-Research system.
Authors: Tomohiro Sawada, Kartik Goyal
Abstract: Standard Byte-Pair Encoding (BPE) tokenization compresses text by pairing a learned token vocabulary with a detailed merge list. Recent work has shown that this merge list exposes a potential attack surface for extracting information about language model's training data. In this paper, we explore the downstream impact of BPE inference algorithms that do not rely on this merge list at all, and hence differ from the encoding process during BPE training. To address this question, we investigate two broad classes of BPE inference schemes that differ from BPE application during training: a) targeted deviation from merge-lists including random merge orders, and various corruptions of merge list involving deletion/truncation, and b) non-targeted BPE inference algorithms that do not depend on the merge list but focus on compressing the text either greedily or exactly. Extensive experiments across diverse language modeling tasks like accuracy-based QA benchmarks, machine translation, and open-ended generation reveal that while targeted deviation from the merge lists exhibits significant degradation in language model performance, the non-targeted merge-list-free inference algorithms result in minimal impact on downstream performance that is often much smaller than expected. These findings pave way for simpler and potentially more privacy-preserving tokenization schemes that do not catastrophically compromise model performance.
Authors: Daniel Wang, Eli Brignac, Minjia Mao, Xiao Fang
Abstract: Large language models (LLMs) are widely applied across diverse domains, raising concerns about their limitations and potential risks. In this study, we investigate two types of bias that LLMs may display: stereotype bias and deviation bias. Stereotype bias refers to when LLMs consistently associate specific traits with a particular demographic group. Deviation bias reflects the disparity between the demographic distributions extracted from LLM-generated content and real-world demographic distributions. By asking four advanced LLMs to generate profiles of individuals, we examine the associations between each demographic group and attributes such as political affiliation, religion, and sexual orientation. Our experimental results show that all examined LLMs exhibit both significant stereotype bias and deviation bias towards multiple groups. Our findings uncover the biases that occur when LLMs infer user attributes and shed light on the potential harms of LLM-generated outputs.
Authors: Jonathan Shaw, Dillon Mee, Timothy Khouw, Zackary Leech, Daniel Wilson
Abstract: Current state-of-the-art models demonstrate capacity to leverage in-context learning to translate into previously unseen language contexts. Tanzer et al. [2024] utilize language materials (e.g. a grammar) to improve translation quality for Kalamang using large language models (LLMs). We focus on Kanuri, a language that, despite having substantial speaker population, has minimal digital resources. We design two datasets for evaluation: one focused on health and humanitarian terms, and another containing generalized terminology, investigating how domain-specific tasks impact LLM translation quality. By providing different combinations of language resources (grammar, dictionary, and parallel sentences), we measure LLM translation effectiveness, comparing results to native speaker translations and human linguist performance. We evaluate using both automatic metrics and native speaker assessments of fluency and accuracy. Results demonstrate that parallel sentences remain the most effective data source, outperforming other methods in human evaluations and automatic metrics. While incorporating grammar improves over zero-shot translation, it fails as an effective standalone data source. Human evaluations reveal that LLMs achieve accuracy (meaning) more effectively than fluency (grammaticality). These findings suggest LLM translation evaluation benefits from multidimensional assessment beyond simple accuracy metrics, and that grammar alone, without parallel sentences, does not provide sufficient context for effective domain-specific translation.
Authors: Swati Rajwal, Shivank Garg, Reem Abdel-Salam, Abdelrahman Zayed
Abstract: The impressive performance of language models is undeniable. However, the presence of biases based on gender, race, socio-economic status, physical appearance, and sexual orientation makes the deployment of language models challenging. This paper studies the effect of chain-of-thought prompting, a recent approach that studies the steps followed by the model before it responds, on fairness. More specifically, we ask the following question: $\textit{Do biased models have biased thoughts}$? To answer our question, we conduct experiments on $5$ popular large language models using fairness metrics to quantify $11$ different biases in the model's thoughts and output. Our results show that the bias in the thinking steps is not highly correlated with the output bias (less than $0.6$ correlation with a $p$-value smaller than $0.001$ in most cases). In other words, unlike human beings, the tested models with biased decisions do not always possess biased thoughts.
Authors: Evangelia Spiliopoulou, Riccardo Fogliato, Hanna Burnsky, Tamer Soliman, Jie Ma, Graham Horwood, Miguel Ballesteros
Abstract: Large language models (LLMs) can serve as judges that offer rapid and reliable assessments of other LLM outputs. However, models may systematically assign overly favorable ratings to their own outputs, a phenomenon known as self-bias, which can distort evaluations of true model performance. Previous studies often conflate genuine differences in model quality with bias or incorrectly assume that evaluations from LLMs and humans follow the same rating distributions. In this work, we present a statistical framework that explicitly formalizes assumptions under which self-bias can be identified and estimated. Our method models the difference in the scoring distribution that LLM-as-a-judge assigns to its own completions compared to other models, while accounting for the underlying quality of the completions provided by an independent, third-party judge (e.g., humans). Our method reliably isolates and quantifies self-bias, even when models vary in ability, ensuring that genuine performance differences are not mistaken for self-bias. We conduct an empirical analysis of self-bias on a large dataset (>5000 prompt-completion pairs) consisting of expert human annotations and judgments from nine different LLM judges. We find that some models, such as GPT-4o and Claude 3.5 Sonnet, systematically assign higher scores to their own outputs. These models also display family-bias; systematically assigning higher ratings to outputs produced by other models of the same family. Our findings highlight potential pitfalls of using LLM judges and offer practical guidance to mitigate biases when interpreting automated evaluations.
Authors: Komala Subramanyam Cherukuri, Pranav Abishai Moses, Aisa Sakata, Jiangping Chen, Haihua Chen
Abstract: Oral histories are vital records of lived experience, particularly within communities affected by systemic injustice and historical erasure. Effective and efficient analysis of their oral history archives can promote access and understanding of the oral histories. However, Large-scale analysis of these archives remains limited due to their unstructured format, emotional complexity, and high annotation costs. This paper presents a scalable framework to automate semantic and sentiment annotation for Japanese American Incarceration Oral History. Using LLMs, we construct a high-quality dataset, evaluate multiple models, and test prompt engineering strategies in historically sensitive contexts. Our multiphase approach combines expert annotation, prompt design, and LLM evaluation with ChatGPT, Llama, and Qwen. We labeled 558 sentences from 15 narrators for sentiment and semantic classification, then evaluated zero-shot, few-shot, and RAG strategies. For semantic classification, ChatGPT achieved the highest F1 score (88.71%), followed by Llama (84.99%) and Qwen (83.72%). For sentiment analysis, Llama slightly outperformed Qwen (82.66%) and ChatGPT (82.29%), with all models showing comparable results. The best prompt configurations were used to annotate 92,191 sentences from 1,002 interviews in the JAIOH collection. Our findings show that LLMs can effectively perform semantic and sentiment annotation across large oral history collections when guided by well-designed prompts. This study provides a reusable annotation pipeline and practical guidance for applying LLMs in culturally sensitive archival analysis. By bridging archival ethics with scalable NLP techniques, this work lays the groundwork for responsible use of artificial intelligence in digital humanities and preservation of collective memory. GitHub: https://github.com/kc6699c/LLM4OralHistoryAnalysis.
Authors: Xianjun Yang, Liqiang Xiao, Shiyang Li, Faisal Ladhak, Hyokun Yun, Linda Ruth Petzold, Yi Xu, William Yang Wang
Abstract: Current jailbreaking work on large language models (LLMs) aims to elicit unsafe outputs from given prompts. However, it only focuses on single-turn jailbreaking targeting one specific query. On the contrary, the advanced LLMs are designed to handle extremely long contexts and can thus conduct multi-turn conversations. So, we propose exploring multi-turn jailbreaking, in which the jailbroken LLMs are continuously tested on more than the first-turn conversation or a single target query. This is an even more serious threat because 1) it is common for users to continue asking relevant follow-up questions to clarify certain jailbroken details, and 2) it is also possible that the initial round of jailbreaking causes the LLMs to respond to additional irrelevant questions consistently. As the first step (First draft done at June 2024) in exploring multi-turn jailbreaking, we construct a Multi-Turn Jailbreak Benchmark (MTJ-Bench) for benchmarking this setting on a series of open- and closed-source models and provide novel insights into this new safety threat. By revealing this new vulnerability, we aim to call for community efforts to build safer LLMs and pave the way for a more in-depth understanding of jailbreaking LLMs.
Authors: Ziqi Liu, Yangbin Chen, Ziyang Zhou, Yilin Li, Mingxuan Hu, Yushan Pan, Zhijie Xu
Abstract: Sarcasm detection is a crucial yet challenging Natural Language Processing task. Existing Large Language Model methods are often limited by single-perspective analysis, static reasoning pathways, and a susceptibility to hallucination when processing complex ironic rhetoric, which impacts their accuracy and reliability. To address these challenges, we propose **SEVADE**, a novel **S**elf-**Ev**olving multi-agent **A**nalysis framework with **D**ecoupled **E**valuation for hallucination-resistant sarcasm detection. The core of our framework is a Dynamic Agentive Reasoning Engine (DARE), which utilizes a team of specialized agents grounded in linguistic theory to perform a multifaceted deconstruction of the text and generate a structured reasoning chain. Subsequently, a separate lightweight rationale adjudicator (RA) performs the final classification based solely on this reasoning chain. This decoupled architecture is designed to mitigate the risk of hallucination by separating complex reasoning from the final judgment. Extensive experiments on four benchmark datasets demonstrate that our framework achieves state-of-the-art performance, with average improvements of **6.75%** in Accuracy and **6.29%** in Macro-F1 score.
Authors: Steven Coyne, Diana Galvan-Sosa, Ryan Spring, Cam\'elia Guerraoui, Michael Zock, Keisuke Sakaguchi, Kentaro Inui
Abstract: Recent advances in natural language processing (NLP) have contributed to the development of automated writing evaluation (AWE) systems that can correct grammatical errors. However, while these systems are effective at improving text, they are not optimally designed for language learning. They favor direct revisions, often with a click-to-fix functionality that can be applied without considering the reason for the correction. Meanwhile, depending on the error type, learners may benefit most from simple explanations and strategically indirect hints, especially on generalizable grammatical rules. To support the generation of such feedback, we introduce an annotation framework that models each error's error type and generalizability. For error type classification, we introduce a typology focused on inferring learners' knowledge gaps by connecting their errors to specific grammatical patterns. Following this framework, we collect a dataset of annotated learner errors and corresponding human-written feedback comments, each labeled as a direct correction or hint. With this data, we evaluate keyword-guided, keyword-free, and template-guided methods of generating feedback using large language models (LLMs). Human teachers examined each system's outputs, assessing them on grounds including relevance, factuality, and comprehensibility. We report on the development of the dataset and the comparative performance of the systems investigated.
Authors: Gangular Singh Irengbam, Nirvash Singh Wahengbam, Lanthoiba Meitei Khumanthem, Paikhomba Oinam
Abstract: This paper presents the development of a Text-to-Speech (TTS) system for the Manipuri language using the Meitei Mayek script. Leveraging Tacotron 2 and HiFi-GAN, we introduce a neural TTS architecture adapted to support tonal phonology and under-resourced linguistic environments. We develop a phoneme mapping for Meitei Mayek to ARPAbet, curate a single-speaker dataset, and demonstrate intelligible and natural speech synthesis, validated through subjective and objective metrics. This system lays the groundwork for linguistic preservation and technological inclusion of Manipuri.
Authors: Xiaobo Zhang (School of Information Engineering, Jiangxi Vocational College of Finance & Economics, Jiujiang, China, School of Computer Sciences, Universiti Sains Malaysia, Penang, Malaysia), Congqing He (School of Computer Sciences, Universiti Sains Malaysia, Penang, Malaysia), Ying He (School of Information Engineering, Jiangxi Vocational College of Finance & Economics, Jiujiang, China, School of Computer Sciences, Universiti Sains Malaysia, Penang, Malaysia), Jian Peng (School of Information Engineering, Jiangxi Vocational College of Finance & Economics, Jiujiang, China), Dajie Fu (School of Information Engineering, Jiangxi Vocational College of Finance & Economics, Jiujiang, China), Tien-Ping Tan (School of Computer Sciences, Universiti Sains Malaysia, Penang, Malaysia)
Abstract: Named Entity Recognition (NER) is a fundamental task in natural language processing. It remains a research hotspot due to its wide applicability across domains. Although recent advances in deep learning have significantly improved NER performance, they rely heavily on large, high-quality annotated datasets. However, building these datasets is expensive and time-consuming, posing a major bottleneck for further research. Current dataset merging approaches mainly focus on strategies like manual label mapping or constructing label graphs, which lack interpretability and scalability. To address this, we propose an automatic label alignment method based on label similarity. The method combines empirical and semantic similarities, using a greedy pairwise merging strategy to unify label spaces across different datasets. Experiments are conducted in two stages: first, merging three existing NER datasets into a unified corpus with minimal impact on NER performance; second, integrating this corpus with a small-scale, self-built dataset in the financial domain. The results show that our method enables effective dataset merging and enhances NER performance in the low-resource financial domain. This study presents an efficient, interpretable, and scalable solution for integrating multi-source NER corpora.
Authors: Philipp Christmann, Gerhard Weikum
Abstract: Personal information is abundant on users' devices, from structured data in calendar, shopping records or fitness tools, to unstructured contents in mail and social media posts. This works presents the ReQAP system that supports users with answers for complex questions that involve filters, joins and aggregation over heterogeneous sources. The unique trait of ReQAP is that it recursively decomposes questions and incrementally builds an operator tree for execution. Both the question interpretation and the individual operators make smart use of light-weight language models, with judicious fine-tuning. The demo showcases the rich functionality for advanced user questions, and also offers detailed tracking of how the answers are computed by the operators in the execution tree. Being able to trace answers back to the underlying sources is vital for human comprehensibility and user trust in the system.
Authors: Arpita Saggar, Jonathan C. Darling, Vania Dimitrova, Duygu Sarikaya, David C. Hogg
Abstract: Persona-based dialogue generation is an important milestone towards building conversational artificial intelligence. Despite the ever-improving capabilities of large language models (LLMs), effectively integrating persona fidelity in conversations remains challenging due to the limited diversity in existing dialogue data. We propose a novel framework SBS (Score-Before-Speaking), which outperforms previous methods and yields improvements for both million and billion-parameter models. Unlike previous methods, SBS unifies the learning of responses and their relative quality into a single step. The key innovation is to train a dialogue model to correlate augmented responses with a quality score during training and then leverage this knowledge at inference. We use noun-based substitution for augmentation and semantic similarity-based scores as a proxy for response quality. Through extensive experiments with benchmark datasets (PERSONA-CHAT and ConvAI2), we show that score-conditioned training allows existing models to better capture a spectrum of persona-consistent dialogues. Our ablation studies also demonstrate that including scores in the input prompt during training is superior to conventional training setups. Code and further details are available at https://arpita2512.github.io/score_before_you_speak
Authors: Siyuan Li, Xi Lin, Guangyan Li, Zehao Liu, Aodu Wulianghai, Li Ding, Jun Wu, Jianhua Li
Abstract: The rapid advancement of large language models (LLMs) has resulted in increasingly sophisticated AI-generated content, posing significant challenges in distinguishing LLM-generated text from human-written language. Existing detection methods, primarily based on lexical heuristics or fine-tuned classifiers, often suffer from limited generalizability and are vulnerable to paraphrasing, adversarial perturbations, and cross-domain shifts. In this work, we propose SentiDetect, a model-agnostic framework for detecting LLM-generated text by analyzing the divergence in sentiment distribution stability. Our method is motivated by the empirical observation that LLM outputs tend to exhibit emotionally consistent patterns, whereas human-written texts display greater emotional variability. To capture this phenomenon, we define two complementary metrics: sentiment distribution consistency and sentiment distribution preservation, which quantify stability under sentiment-altering and semantic-preserving transformations. We evaluate SentiDetect on five diverse datasets and a range of advanced LLMs,including Gemini-1.5-Pro, Claude-3, GPT-4-0613, and LLaMa-3.3. Experimental results demonstrate its superiority over state-of-the-art baselines, with over 16% and 11% F1 score improvements on Gemini-1.5-Pro and GPT-4-0613, respectively. Moreover, SentiDetect also shows greater robustness to paraphrasing, adversarial attacks, and text length variations, outperforming existing detectors in challenging scenarios.
Authors: Mohamed Basem, Islam Oshallah, Ali Hamdi, Khaled Shaban, Hozaifa Kassab
Abstract: Quranic Question Answering presents unique challenges due to the linguistic complexity of Classical Arabic and the semantic richness of religious texts. In this paper, we propose a novel two-stage framework that addresses both passage retrieval and answer extraction. For passage retrieval, we ensemble fine-tuned Arabic language models to achieve superior ranking performance. For answer extraction, we employ instruction-tuned large language models with few-shot prompting to overcome the limitations of fine-tuning on small datasets. Our approach achieves state-of-the-art results on the Quran QA 2023 Shared Task, with a MAP@10 of 0.3128 and MRR@10 of 0.5763 for retrieval, and a pAP@10 of 0.669 for extraction, substantially outperforming previous methods. These results demonstrate that combining model ensembling and instruction-tuned language models effectively addresses the challenges of low-resource question answering in specialized domains.
Authors: Zhijun Tu, Hanting Chen, Siqi Liu, Chuanjian Liu, Jian Li, Jie Hu, Yunhe Wang
Abstract: 1-bit LLM quantization offers significant advantages in reducing storage and computational costs. However, existing methods typically train 1-bit LLMs from scratch, failing to fully leverage pre-trained models. This results in high training costs and notable accuracy degradation. We identify that the large gap between full precision and 1-bit representations makes direct adaptation difficult. In this paper, we introduce a consistent progressive training for both forward and backward, smoothly converting the floating-point weights into the binarized ones. Additionally, we incorporate binary-aware initialization and dual-scaling compensation to reduce the difficulty of progressive training and improve the performance. Experimental results on LLMs of various sizes demonstrate that our method outperforms existing approaches. Our results show that high-performance 1-bit LLMs can be achieved using pre-trained models, eliminating the need for expensive training from scratch.
Authors: Mao Li, Fred Conrad, Johann Gagnon-Bartsch
Abstract: We propose Vec2Summ, a novel method for abstractive summarization that frames the task as semantic compression. Vec2Summ represents a document collection using a single mean vector in the semantic embedding space, capturing the central meaning of the corpus. To reconstruct fluent summaries, we perform embedding inversion -- decoding this mean vector into natural language using a generative language model. To improve reconstruction quality and capture some degree of topical variability, we introduce stochasticity by sampling from a Gaussian distribution centered on the mean. This approach is loosely analogous to bagging in ensemble learning, where controlled randomness encourages more robust and varied outputs. Vec2Summ addresses key limitations of LLM-based summarization methods. It avoids context-length constraints, enables interpretable and controllable generation via semantic parameters, and scales efficiently with corpus size -- requiring only $O(d + d^2)$ parameters. Empirical results show that Vec2Summ produces coherent summaries for topically focused, order-invariant corpora, with performance comparable to direct LLM summarization in terms of thematic coverage and efficiency, albeit with less fine-grained detail. These results underscore Vec2Summ's potential in settings where scalability, semantic control, and corpus-level abstraction are prioritized.
Authors: Muhammad Dehan Al Kautsar, Aswin Candra, Muhammad Alif Al Hakim, Maxalmina Satria Kahfi, Fajri Koto, Alham Fikri Aji, Peerat Limkonchotiwat, Ekapol Chuangsuwanich, Genta Indra Winata
Abstract: Although numerous datasets have been developed to support dialogue systems, most existing chit-chat datasets overlook the cultural nuances inherent in natural human conversations. To address this gap, we introduce SEADialogues, a culturally grounded dialogue dataset centered on Southeast Asia, a region with over 700 million people and immense cultural diversity. Our dataset features dialogues in eight languages from six Southeast Asian countries, many of which are low-resource despite having sizable speaker populations. To enhance cultural relevance and personalization, each dialogue includes persona attributes and two culturally grounded topics that reflect everyday life in the respective communities. Furthermore, we release a multi-turn dialogue dataset to advance research on culturally aware and human-centric large language models, including conversational dialogue agents.
Authors: Aditya Tomar, Nihar Ranjan Sahoo, Pushpak Bhattacharyya
Abstract: Evaluating social biases in language models (LMs) is crucial for ensuring fairness and minimizing the reinforcement of harmful stereotypes in AI systems. Existing benchmarks, such as the Bias Benchmark for Question Answering (BBQ), primarily focus on Western contexts, limiting their applicability to the Indian context. To address this gap, we introduce BharatBBQ, a culturally adapted benchmark designed to assess biases in Hindi, English, Marathi, Bengali, Tamil, Telugu, Odia, and Assamese. BharatBBQ covers 13 social categories, including 3 intersectional groups, reflecting prevalent biases in the Indian sociocultural landscape. Our dataset contains 49,108 examples in one language that are expanded using translation and verification to 392,864 examples in eight different languages. We evaluate five multilingual LM families across zero and few-shot settings, analyzing their bias and stereotypical bias scores. Our findings highlight persistent biases across languages and social categories and often amplified biases in Indian languages compared to English, demonstrating the necessity of linguistically and culturally grounded benchmarks for bias evaluation.
Authors: Lijie Yang, Zhihao Zhang, Arti Jain, Shijie Cao, Baihong Yuan, Yiwei Chen, Zhihao Jia, Ravi Netravali
Abstract: Large reasoning models achieve strong performance through test-time scaling but incur substantial computational overhead, particularly from excessive token generation when processing short input prompts. While sparse attention mechanisms can reduce latency and memory usage, existing approaches suffer from significant accuracy degradation due to accumulated errors during long-generation reasoning. These methods generally require either high token retention rates or expensive retraining. We introduce LessIsMore, a training-free sparse attention mechanism for reasoning tasks, which leverages global attention patterns rather than relying on traditional head-specific local optimizations. LessIsMore aggregates token selections from local attention heads with recent contextual information, enabling unified cross-head token ranking for future decoding layers. This unified selection improves generalization and efficiency by avoiding the need to maintain separate token subsets per head. Evaluation across diverse reasoning tasks and benchmarks shows that LessIsMore preserves -- and in some cases improves -- accuracy while achieving a $1.1\times$ average decoding speed-up compared to full attention. Moreover, LessIsMore attends to $2\times$ fewer tokens without accuracy loss, achieving a $1.13\times$ end-to-end speed-up compared to existing sparse attention methods.
Authors: Falaah Arif Khan, Nivedha Sivakumar, Yinong Oliver Wang, Katherine Metcalf, Cezanne Camacho, Barry-John Theobald, Luca Zappella, Nicholas Apostoloff
Abstract: Large language models (LLMs) have achieved impressive performance, leading to their widespread adoption as decision-support tools in resource-constrained contexts like hiring and admissions. There is, however, scientific consensus that AI systems can reflect and exacerbate societal biases, raising concerns about identity-based harm when used in critical social contexts. Prior work has laid a solid foundation for assessing bias in LLMs by evaluating demographic disparities in different language reasoning tasks. In this work, we extend single-axis fairness evaluations to examine intersectional bias, recognizing that when multiple axes of discrimination intersect, they create distinct patterns of disadvantage. We create a new benchmark called WinoIdentity by augmenting the WinoBias dataset with 25 demographic markers across 10 attributes, including age, nationality, and race, intersected with binary gender, yielding 245,700 prompts to evaluate 50 distinct bias patterns. Focusing on harms of omission due to underrepresentation, we investigate bias through the lens of uncertainty and propose a group (un)fairness metric called Coreference Confidence Disparity which measures whether models are more or less confident for some intersectional identities than others. We evaluate five recently published LLMs and find confidence disparities as high as 40% along various demographic attributes including body type, sexual orientation and socio-economic status, with models being most uncertain about doubly-disadvantaged identities in anti-stereotypical settings. Surprisingly, coreference confidence decreases even for hegemonic or privileged markers, indicating that the recent impressive performance of LLMs is more likely due to memorization than logical reasoning. Notably, these are two independent failures in value alignment and validity that can compound to cause social harm.
Authors: Anna Seo Gyeong Choi, Hoon Choi
Abstract: Automatic Speech Recognition (ASR) systems now mediate countless human-technology interactions, yet research on their fairness implications remains surprisingly limited. This paper examines ASR bias through a philosophical lens, arguing that systematic misrecognition of certain speech varieties constitutes more than a technical limitation -- it represents a form of disrespect that compounds historical injustices against marginalized linguistic communities. We distinguish between morally neutral classification (discriminate1) and harmful discrimination (discriminate2), demonstrating how ASR systems can inadvertently transform the former into the latter when they consistently misrecognize non-standard dialects. We identify three unique ethical dimensions of speech technologies that differentiate ASR bias from other algorithmic fairness concerns: the temporal burden placed on speakers of non-standard varieties ("temporal taxation"), the disruption of conversational flow when systems misrecognize speech, and the fundamental connection between speech patterns and personal/cultural identity. These factors create asymmetric power relationships that existing technical fairness metrics fail to capture. The paper analyzes the tension between linguistic standardization and pluralism in ASR development, arguing that current approaches often embed and reinforce problematic language ideologies. We conclude that addressing ASR bias requires more than technical interventions; it demands recognition of diverse speech varieties as legitimate forms of expression worthy of technological accommodation. This philosophical reframing offers new pathways for developing ASR systems that respect linguistic diversity and speaker autonomy.
Authors: Biao Yi, Jiahao Li, Baolei Zhang, Lihai Nie, Tong Li, Tiansheng Huang, Zheli Liu
Abstract: Fine-tuning-as-a-Service introduces a critical vulnerability where a few malicious examples mixed into the user's fine-tuning dataset can compromise the safety alignment of Large Language Models (LLMs). While a recognized paradigm frames safe fine-tuning as a multi-objective optimization problem balancing user task performance with safety alignment, we find existing solutions are critically sensitive to the harmful ratio, with defenses degrading sharply as harmful ratio increases. We diagnose that this failure stems from conflicting gradients, where the user-task update directly undermines the safety objective. To resolve this, we propose SafeGrad, a novel method that employs gradient surgery. When a conflict is detected, SafeGrad nullifies the harmful component of the user-task gradient by projecting it onto the orthogonal plane of the alignment gradient, allowing the model to learn the user's task without sacrificing safety. To further enhance robustness and data efficiency, we employ a KL-divergence alignment loss that learns the rich, distributional safety profile of the well-aligned foundation model. Extensive experiments show that SafeGrad provides state-of-the-art defense across various LLMs and datasets, maintaining robust safety even at high harmful ratios without compromising task fidelity.
Authors: Leyi Pan, Zheyu Fu, Yunpeng Zhai, Shuchang Tao, Sheng Guan, Shiyu Huang, Lingzhe Zhang, Zhaoyang Liu, Bolin Ding, Felix Henry, Lijie Wen, Aiwei Liu
Abstract: The rise of Omni-modal Large Language Models (OLLMs), which integrate visual and auditory processing with text, necessitates robust safety evaluations to mitigate harmful outputs. However, no dedicated benchmarks currently exist for OLLMs, and prior benchmarks designed for other LLMs lack the ability to assess safety performance under audio-visual joint inputs or cross-modal safety consistency. To fill this gap, we introduce Omni-SafetyBench, the first comprehensive parallel benchmark for OLLM safety evaluation, featuring 24 modality combinations and variations with 972 samples each, including dedicated audio-visual harm cases. Considering OLLMs' comprehension challenges with complex omni-modal inputs and the need for cross-modal consistency evaluation, we propose tailored metrics: a Safety-score based on conditional Attack Success Rate (C-ASR) and Refusal Rate (C-RR) to account for comprehension failures, and a Cross-Modal Safety Consistency Score (CMSC-score) to measure consistency across modalities. Evaluating 6 open-source and 4 closed-source OLLMs reveals critical vulnerabilities: (1) no model excels in both overall safety and consistency, with only 3 models achieving over 0.6 in both metrics and top performer scoring around 0.8; (2) safety defenses weaken with complex inputs, especially audio-visual joints; (3) severe weaknesses persist, with some models scoring as low as 0.14 on specific modalities. Our benchmark and metrics highlight urgent needs for enhanced OLLM safety, providing a foundation for future improvements.
Authors: Kejin Liu, Junhong Lian, Xiang Ao, Ningtao Wang, Xing Fu, Yu Cheng, Weiqiang Wang, Xinyu Liu
Abstract: Accurate personalized headline generation hinges on precisely capturing user interests from historical behaviors. However, existing methods neglect personalized-irrelevant click noise in entire historical clickstreams, which may lead to hallucinated headlines that deviate from genuine user preferences. In this paper, we reveal the detrimental impact of click noise on personalized generation quality through rigorous analysis in both user and news dimensions. Based on these insights, we propose a novel Personalized Headline Generation framework via Denoising Fake Interests from Implicit Feedback (PHG-DIF). PHG-DIF first employs dual-stage filtering to effectively remove clickstream noise, identified by short dwell times and abnormal click bursts, and then leverages multi-level temporal fusion to dynamically model users' evolving and multi-faceted interests for precise profiling. Moreover, we release DT-PENS, a new benchmark dataset comprising the click behavior of 1,000 carefully curated users and nearly 10,000 annotated personalized headlines with historical dwell time annotations. Extensive experiments demonstrate that PHG-DIF substantially mitigates the adverse effects of click noise and significantly improves headline quality, achieving state-of-the-art (SOTA) results on DT-PENS. Our framework implementation and dataset are available at https://github.com/liukejin-up/PHG-DIF.
Authors: Jiaqi Yin, Yi-Wei Chen, Meng-Lung Lee, Xiya Liu
Abstract: Enterprise data pipelines, characterized by complex transformations across multiple programming languages, often cause a semantic disconnect between original metadata and downstream data. This "semantic drift" compromises data reproducibility and governance, and impairs the utility of services like retrieval-augmented generation (RAG) and text-to-SQL systems. To address this, a novel framework is proposed for the automated extraction of fine-grained schema lineage from multilingual enterprise pipeline scripts. This method identifies four key components: source schemas, source tables, transformation logic, and aggregation operations, creating a standardized representation of data transformations. For the rigorous evaluation of lineage quality, this paper introduces the Schema Lineage Composite Evaluation (SLiCE), a metric that assesses both structural correctness and semantic fidelity. A new benchmark is also presented, comprising 1,700 manually annotated lineages from real-world industrial scripts. Experiments were conducted with 12 language models, from 1.3B to 32B small language models (SLMs) to large language models (LLMs) like GPT-4o and GPT-4.1. The results demonstrate that the performance of schema lineage extraction scales with model size and the sophistication of prompting techniques. Specially, a 32B open-source model, using a single reasoning trace, can achieve performance comparable to the GPT series under standard prompting. This finding suggests a scalable and economical approach for deploying schema-aware agents in practical applications.
Authors: Kabir Khan, Priya Sharma, Arjun Mehta, Neha Gupta, Ravi Narayanan
Abstract: Large Language Models (LLMs) suffer from a critical limitation: their knowledge is static and quickly becomes outdated. Retraining these massive models is computationally prohibitive, while existing knowledge editing techniques can be slow and may introduce unforeseen side effects. To address this, we propose DySK-Attn, a novel framework that enables LLMs to efficiently integrate real-time knowledge from a dynamic external source. Our approach synergizes an LLM with a dynamic Knowledge Graph (KG) that can be updated instantaneously. The core of our framework is a sparse knowledge attention mechanism, which allows the LLM to perform a coarse-to-fine grained search, efficiently identifying and focusing on a small, highly relevant subset of facts from the vast KG. This mechanism avoids the high computational cost of dense attention over the entire knowledge base and mitigates noise from irrelevant information. We demonstrate through extensive experiments on time-sensitive question-answering tasks that DySK-Attn significantly outperforms strong baselines, including standard Retrieval-Augmented Generation (RAG) and model editing techniques, in both factual accuracy for updated knowledge and computational efficiency. Our framework offers a scalable and effective solution for building LLMs that can stay current with the ever-changing world.
Authors: Yanru Sun, Emadeldeen Eldele, Zongxia Xie, Yucheng Wang, Wenzhe Niu, Qinghua Hu, Chee Keong Kwoh, Min Wu
Abstract: Large Language Models (LLMs) have recently demonstrated impressive capabilities in natural language processing due to their strong generalization and sequence modeling capabilities. However, their direct application to time series forecasting remains challenging due to two fundamental issues: the inherent heterogeneity of temporal patterns and the modality gap between continuous numerical signals and discrete language representations. In this work, we propose TALON, a unified framework that enhances LLM-based forecasting by modeling temporal heterogeneity and enforcing semantic alignment. Specifically, we design a Heterogeneous Temporal Encoder that partitions multivariate time series into structurally coherent segments, enabling localized expert modeling across diverse temporal patterns. To bridge the modality gap, we introduce a Semantic Alignment Module that aligns temporal features with LLM-compatible representations, enabling effective integration of time series into language-based models while eliminating the need for handcrafted prompts during inference. Extensive experiments on seven real-world benchmarks demonstrate that TALON achieves superior performance across all datasets, with average MSE improvements of up to 11\% over recent state-of-the-art methods. These results underscore the effectiveness of incorporating both pattern-aware and semantic-aware designs when adapting LLMs for time series forecasting. The code is available at: https://github.com/syrGitHub/TALON.
Authors: Chaoqun Cui, Siyuan Li, Kunkun Ma, Caiyan Jia
Abstract: Pretrained Language Models (PLMs) have excelled in various Natural Language Processing tasks, benefiting from large-scale pretraining and self-attention mechanism's ability to capture long-range dependencies. However, their performance on social media application tasks like rumor detection remains suboptimal. We attribute this to mismatches between pretraining corpora and social texts, inadequate handling of unique social symbols, and pretraining tasks ill-suited for modeling user engagements implicit in propagation structures. To address these issues, we propose a continue pretraining strategy called Post Engagement Prediction (PEP) to infuse information from propagation structures into PLMs. PEP makes models to predict root, branch, and parent relations between posts, capturing interactions of stance and sentiment crucial for rumor detection. We also curate and release large-scale Twitter corpus: TwitterCorpus (269GB text), and two unlabeled claim conversation datasets with propagation structures (UTwitter and UWeibo). Utilizing these resources and PEP strategy, we train a Twitter-tailored PLM called SoLM. Extensive experiments demonstrate PEP significantly boosts rumor detection performance across universal and social media PLMs, even in few-shot scenarios. On benchmark datasets, PEP enhances baseline models by 1.0-3.7\% accuracy, even enabling it to outperform current state-of-the-art methods on multiple datasets. SoLM alone, without high-level modules, also achieves competitive results, highlighting the strategy's effectiveness in learning discriminative post interaction features.
Authors: Itai Allouche, Itay Asael, Rotem Rousso, Vered Dassa, Ann Bradlow, Seung-Eun Kim, Matthew Goldrick, Joseph Keshet
Abstract: Despite their success in speech processing, neural networks often operate as black boxes, prompting the question: what informs their decisions, and how can we interpret them? This work examines this issue in the context of lexical stress. A dataset of English disyllabic words was automatically constructed from read and spontaneous speech. Several Convolutional Neural Network (CNN) architectures were trained to predict stress position from a spectrographic representation of disyllabic words lacking minimal stress pairs (e.g., initial stress WAllet, final stress exTEND), achieving up to 92% accuracy on held-out test data. Layerwise Relevance Propagation (LRP), a technique for CNN interpretability analysis, revealed that predictions for held-out minimal pairs (PROtest vs. proTEST ) were most strongly influenced by information in stressed versus unstressed syllables, particularly the spectral properties of stressed vowels. However, the classifiers also attended to information throughout the word. A feature-specific relevance analysis is proposed, and its results suggest that our best-performing classifier is strongly influenced by the stressed vowel's first and second formants, with some evidence that its pitch and third formant also contribute. These results reveal deep learning's ability to acquire distributed cues to stress from naturally occurring data, extending traditional phonetic work based around highly controlled stimuli.
Authors: Zhe Ren
Abstract: Knowledge distillation has been successfully applied to Continual Learning Named Entity Recognition (CLNER) tasks, by using a teacher model trained on old-class data to distill old-class entities present in new-class data as a form of regularization, thereby avoiding catastrophic forgetting. However, in Few-Shot CLNER (FS-CLNER) tasks, the scarcity of new-class entities makes it difficult for the trained model to generalize during inference. More critically, the lack of old-class entity information hinders the distillation of old knowledge, causing the model to fall into what we refer to as the Few-Shot Distillation Dilemma. In this work, we address the above challenges through a prompt tuning paradigm and memory demonstration template strategy. Specifically, we designed an expandable Anchor words-oriented Prompt Tuning (APT) paradigm to bridge the gap between pre-training and fine-tuning, thereby enhancing performance in few-shot scenarios. Additionally, we incorporated Memory Demonstration Templates (MDT) into each training instance to provide replay samples from previous tasks, which not only avoids the Few-Shot Distillation Dilemma but also promotes in-context learning. Experiments show that our approach achieves competitive performances on FS-CLNER.
Authors: Bernd J. Kr\"oger
Abstract: This paper describes an extension of the two-dimensional dynamic articulatory model DYNARTmo by integrating an internal three-dimensional representation of the palatal dome to estimate tongue-palate contact areas from midsagittal tongue contours. Two alternative dome geometries - a half-ellipse and a cosine based profile - are implemented to model lateral curvature in the coronal plane. Using these geometries, lateral contact points are analytically computed for each anterior-posterior position, enabling the generation of electropalatography-like visualizations within the 2D+ framework. The enhanced model supports three synchronized views (sagittal, glottal, and palatal) for static and dynamic (animated) articulation displays, suitable for speech science education and speech therapy. Future work includes adding a facial (lip) view and implementing articulatory-to-acoustic synthesis to quantitatively evaluate model realism.
Authors: Qiongqiong Wang, Hardik B. Sailor, Jeremy H. M. Wong, Tianchi Liu, Shuo Sun, Wenyu Zhang, Muhammad Huzaifah, Nancy Chen, Ai Ti Aw
Abstract: Current large speech language models (Speech-LLMs) often exhibit limitations in empathetic reasoning, primarily due to the absence of training datasets that integrate both contextual content and paralinguistic cues. In this work, we propose two approaches to incorporate contextual paralinguistic information into model training: (1) an explicit method that provides paralinguistic metadata (e.g., emotion annotations) directly to the LLM, and (2) an implicit method that automatically generates novel training question-answer (QA) pairs using both categorical and dimensional emotion annotations alongside speech transcriptions. Our implicit method boosts performance (LLM-judged) by 38.41% on a human-annotated QA benchmark, reaching 46.02% when combined with the explicit approach, showing effectiveness in contextual paralinguistic understanding. We also validate the LLM judge by demonstrating its correlation with classification metrics, providing support for its reliability.
Authors: Vasudha Varadarajan, Hui Xu, Rebecca Astrid Boehme, Mariam Marlan Mirstrom, Sverker Sikstrom, H. Andrew Schwartz
Abstract: Recent advances in large language models (LLMs) offer new opportunities for scalable, interactive mental health assessment, but excessive querying by LLMs burdens users and is inefficient for real-world screening across transdiagnostic symptom profiles. We introduce MAQuA, an adaptive question-asking framework for simultaneous, multidimensional mental health screening. Combining multi-outcome modeling on language responses with item response theory (IRT) and factor analysis, MAQuA selects the questions with most informative responses across multiple dimensions at each turn to optimize diagnostic information, improving accuracy and potentially reducing response burden. Empirical results on a novel dataset reveal that MAQuA reduces the number of assessment questions required for score stabilization by 50-87% compared to random ordering (e.g., achieving stable depression scores with 71% fewer questions and eating disorder scores with 85% fewer questions). MAQuA demonstrates robust performance across both internalizing (depression, anxiety) and externalizing (substance use, eating disorder) domains, with early stopping strategies further reducing patient time and burden. These findings position MAQuA as a powerful and efficient tool for scalable, nuanced, and interactive mental health screening, advancing the integration of LLM-based agents into real-world clinical workflows.
Authors: Junchen Ding, Penghao Jiang, Zihao Xu, Ziqi Ding, Yichen Zhu, Jiaojiao Jiang, Yuekang Li
Abstract: As large language models (LLMs) increasingly mediate ethically sensitive decisions, understanding their moral reasoning processes becomes imperative. This study presents a comprehensive empirical evaluation of 14 leading LLMs, both reasoning enabled and general purpose, across 27 diverse trolley problem scenarios, framed by ten moral philosophies, including utilitarianism, deontology, and altruism. Using a factorial prompting protocol, we elicited 3,780 binary decisions and natural language justifications, enabling analysis along axes of decisional assertiveness, explanation answer consistency, public moral alignment, and sensitivity to ethically irrelevant cues. Our findings reveal significant variability across ethical frames and model types: reasoning enhanced models demonstrate greater decisiveness and structured justifications, yet do not always align better with human consensus. Notably, "sweet zones" emerge in altruistic, fairness, and virtue ethics framings, where models achieve a balance of high intervention rates, low explanation conflict, and minimal divergence from aggregated human judgments. However, models diverge under frames emphasizing kinship, legality, or self interest, often producing ethically controversial outcomes. These patterns suggest that moral prompting is not only a behavioral modifier but also a diagnostic tool for uncovering latent alignment philosophies across providers. We advocate for moral reasoning to become a primary axis in LLM alignment, calling for standardized benchmarks that evaluate not just what LLMs decide, but how and why.
Authors: Jian Chen, Jinbao Tian, Yankui Li, Zhou Li
Abstract: Accurate information extraction from specialized texts is a critical challenge, particularly for named entity recognition (NER) in the architecture, engineering, and construction (AEC) domain to support automated rule checking (ARC). The performance of standard pre-trained models is often constrained by the domain gap, as they struggle to interpret the specialized terminology and complex relational contexts inherent in AEC texts. Although this issue can be mitigated by further pre-training on large, human-curated domain corpora, as exemplified by methods like ARCBERT, this approach is both labor-intensive and cost-prohibitive. Consequently, leveraging large language models (LLMs) for automated knowledge generation has emerged as a promising alternative. However, the optimal strategy for generating knowledge that can genuinely enhance smaller, efficient models remains an open question. To address this, we propose ARCE (augmented RoBERTa with contextualized elucidations), a novel approach that systematically explores and optimizes this generation process. ARCE employs an LLM to first generate a corpus of simple, direct explanations, which we term Cote, and then uses this corpus to incrementally pre-train a RoBERTa model prior to its fine-tuning on the downstream task. Our extensive experiments show that ARCE establishes a new state-of-the-art on a benchmark AEC dataset, achieving a Macro-F1 score of 77.20%. This result also reveals a key finding: simple, explanation-based knowledge proves surprisingly more effective than complex, role-based rationales for this task. The code is publicly available at:https://github.com/nxcc-lab/ARCE.
Authors: Yexing Du, Kaiyuan Liu, Youcheng Pan, Zheng Chu, Bo Yang, Xiaocheng Feng, Yang Xiang, Ming Liu
Abstract: As Large Language Models (LLMs) are increasingly popularized in the multilingual world, ensuring hallucination-free factuality becomes markedly crucial. However, existing benchmarks for evaluating the reliability of Multimodal Large Language Models (MLLMs) predominantly focus on textual or visual modalities with a primary emphasis on English, which creates a gap in evaluation when processing multilingual input, especially in speech. To bridge this gap, we propose a novel \textbf{C}ross-lingual and \textbf{C}ross-modal \textbf{F}actuality benchmark (\textbf{CCFQA}). Specifically, the CCFQA benchmark contains parallel speech-text factual questions across 8 languages, designed to systematically evaluate MLLMs' cross-lingual and cross-modal factuality capabilities. Our experimental results demonstrate that current MLLMs still face substantial challenges on the CCFQA benchmark. Furthermore, we propose a few-shot transfer learning strategy that effectively transfers the Question Answering (QA) capabilities of LLMs in English to multilingual Spoken Question Answering (SQA) tasks, achieving competitive performance with GPT-4o-mini-Audio using just 5-shot training. We release CCFQA as a foundational research resource to promote the development of MLLMs with more robust and reliable speech understanding capabilities. Our code and dataset are available at https://github.com/yxduir/ccfqa.
Authors: Cristian Cosentino, Annamaria Defilippo, Marco Dossena, Christopher Irwin, Sara Joubbi, Pietro Li\`o
Abstract: HealthBranches is a novel benchmark dataset for medical Question-Answering (Q&A), specifically designed to evaluate complex reasoning in Large Language Models (LLMs). This dataset is generated through a semi-automated pipeline that transforms explicit decision pathways from medical source into realistic patient cases with associated questions and answers. Covering 4,063 case studies across 17 healthcare topics, each data point is based on clinically validated reasoning chains. HealthBranches supports both open-ended and multiple-choice question formats and uniquely includes the full reasoning path for each Q&A. Its structured design enables robust evaluation of LLMs' multi-step inference capabilities, including their performance in structured Retrieval-Augmented Generation (RAG) contexts. HealthBranches establishes a foundation for the development of more trustworthy, interpretable, and clinically reliable LLMs in high-stakes domains while also serving as a valuable resource for educational purposes.
Authors: Shubhra Ghosh, Abhilekh Borah, Aditya Kumar Guru, Kripabandhu Ghosh
Abstract: The rapid proliferation of Large Language Models (LLMs) has significantly contributed to the development of equitable AI systems capable of factual question-answering (QA). However, no known study tests the LLMs' robustness when presented with obfuscated versions of questions. To systematically evaluate these limitations, we propose a novel technique, ObfusQAte and, leveraging the same, introduce ObfusQA, a comprehensive, first of its kind, framework with multi-tiered obfuscation levels designed to examine LLM capabilities across three distinct dimensions: (i) Named-Entity Indirection, (ii) Distractor Indirection, and (iii) Contextual Overload. By capturing these fine-grained distinctions in language, ObfusQA provides a comprehensive benchmark for evaluating LLM robustness and adaptability. Our study observes that LLMs exhibit a tendency to fail or generate hallucinated responses when confronted with these increasingly nuanced variations. To foster research in this direction, we make ObfusQAte publicly available.
Authors: Dean Geckt, Melinda Fricke, Shuly Wintner
Abstract: Most people are multilingual, and most multilinguals code-switch, yet the characteristics of code-switched language are not fully understood. We developed a chatbot capable of completing a Map Task with human participants using code-switched Spanish and English. In two experiments, we prompted the bot to code-switch according to different strategies, examining (1) the feasibility of such experiments for investigating bilingual language use, and (2) whether participants would be sensitive to variations in discourse and grammatical patterns. Participants generally enjoyed code-switching with our bot as long as it produced predictable code-switching behavior; when code-switching was random or ungrammatical (as when producing unattested incongruent mixed-language noun phrases, such as `la fork'), participants enjoyed the task less and were less successful at completing it. These results underscore the potential downsides of deploying insufficiently developed multilingual language technology, while also illustrating the promise of such technology for conducting research on bilingual language use.
Authors: Wenqian Cui, Lei Zhu, Xiaohui Li, Zhihan Guo, Haoli Bai, Lu Hou, Irwin King
Abstract: Full-Duplex Speech Language Models (FD-SLMs) are specialized foundation models designed to enable natural, real-time spoken interactions by modeling complex conversational dynamics such as interruptions, backchannels, and overlapping speech, and End-to-end (e2e) FD-SLMs leverage real-world double-channel conversational data to capture nuanced two-speaker dialogue patterns for human-like interactions. However, they face a critical challenge -- their conversational abilities often degrade compared to pure-text conversation due to prolonged speech sequences and limited high-quality spoken dialogue data. While text-guided speech generation could mitigate these issues, it suffers from timing and length issues when integrating textual guidance into double-channel audio streams, disrupting the precise time alignment essential for natural interactions. To address these challenges, we propose TurnGuide, a novel planning-inspired approach that mimics human conversational planning by dynamically segmenting assistant speech into dialogue turns and generating turn-level text guidance before speech output, which effectively resolves both insertion timing and length challenges. Extensive experiments demonstrate our approach significantly improves e2e FD-SLMs' conversational abilities, enabling them to generate semantically meaningful and coherent speech while maintaining natural conversational flow. Demos are available at https://dreamtheater123.github.io/TurnGuide-Demo/. Code will be available at https://github.com/dreamtheater123/TurnGuide.
URLs: https://dreamtheater123.github.io/TurnGuide-Demo/., https://github.com/dreamtheater123/TurnGuide.
Authors: Jean de Dieu Nyandwi, Yueqi Song, Simran Khanuja, Graham Neubig
Abstract: Multimodal Large Language Models excel in high-resource settings, but often misinterpret long-tail cultural entities and underperform in low-resource languages. To address this gap, we propose a data-centric approach that directly grounds MLLMs in cultural knowledge. Leveraging a large scale knowledge graph from Wikidata, we collect images that represent culturally significant entities, and generate synthetic multilingual visual question answering data. The resulting dataset, CulturalGround, comprises 22 million high-quality, culturally-rich VQA pairs spanning 42 countries and 39 languages. We train an open-source MLLM CulturalPangea on CulturalGround, interleaving standard multilingual instruction-tuning data to preserve general abilities. CulturalPangea achieves state-of-the-art performance among open models on various culture-focused multilingual multimodal benchmarks, outperforming prior models by an average of 5.0 without degrading results on mainstream vision-language tasks. Our findings show that our targeted, culturally grounded approach could substantially narrow the cultural gap in MLLMs and offer a practical path towards globally inclusive multimodal systems.
Authors: Zhiyi Lyu, Jianguo Huang, Yanchen Deng, Steven Hoi, Bo An
Abstract: Large Language Models (LLMs) with inference-time scaling techniques show promise for code generation, yet face notable efficiency and scalability challenges. Construction-based tree-search methods suffer from rapid growth in tree size, high token consumption, and lack of anytime property. In contrast, improvement-based methods offer better performance but often struggle with uninformative reward signals and inefficient search strategies. In this work, we propose \textbf{ReLoc}, a unified local search framework which effectively performs step-by-step code revision. Specifically, ReLoc explores a series of local revisions through four key algorithmic components: initial code drafting, neighborhood code generation, candidate evaluation, and incumbent code updating, each of which can be instantiated with specific decision rules to realize different local search algorithms such as Hill Climbing (HC) or Genetic Algorithm (GA). Furthermore, we develop a specialized revision reward model that evaluates code quality based on revision distance to produce fine-grained preferences that guide the local search toward more promising candidates. Finally, our extensive experimental results demonstrate that our approach achieves superior performance across diverse code generation tasks, significantly outperforming both construction-based tree search as well as the state-of-the-art improvement-based code generation methods.
Authors: Blerta Veseli, Julian Chibane, Mariya Toneva, Alexander Koller
Abstract: Large Language Models (LLMs) often struggle to use information across long inputs effectively. Prior work has identified positional biases, such as the Lost in the Middle (LiM) effect, where models perform better when information appears at the beginning (primacy bias) or end (recency bias) of the input, rather than in the middle. However, long-context studies have not consistently replicated these effects, raising questions about their intensity and the conditions under which they manifest. To address this, we conducted a comprehensive analysis using relative rather than absolute input lengths, defined with respect to each model's context window. Our findings reveal that the LiM effect is strongest when inputs occupy up to 50% of a model's context window. Beyond that, the primacy bias weakens, while recency bias remains relatively stable. This effectively eliminates the LiM effect; instead, we observe a distance-based bias, where model performance is better when relevant information is closer to the end of the input. Furthermore, our results suggest that successful retrieval is a prerequisite for reasoning in LLMs, and that the observed positional biases in reasoning are largely inherited from retrieval. These insights have implications for long-context tasks, the design of future LLM benchmarks, and evaluation methodologies for LLMs handling extended inputs.
Authors: Archchana Sindhujan, Shenbin Qian, Chan Chi Chun Matthew, Constantin Orasan, Diptesh Kanojia
Abstract: Large Language Models (LLMs) have shown remarkable performance across a wide range of natural language processing tasks. Quality Estimation (QE) for Machine Translation (MT), which assesses the quality of a source-target pair without relying on reference translations, remains a challenging cross-lingual task for LLMs. The challenges stem from the inherent limitations of existing LLM-based QE systems, which are pre-trained for causal language modelling rather than regression-specific tasks, further elevated by the presence of low-resource languages given pre-training data distribution. This paper introduces ALOPE, an adaptive layer-optimization framework designed to enhance LLM-based QE by restructuring Transformer representations through layer-wise adaptation for improved regression-based prediction. Our framework integrates low-rank adapters (LoRA) with regression task heads, leveraging selected pre-trained Transformer layers for improved cross-lingual alignment. In addition to the layer-specific adaptation, ALOPE introduces two strategies-dynamic weighting, which adaptively combines representations from multiple layers, and multi-head regression, which aggregates regression losses from multiple heads for QE. Our framework shows improvements over various existing LLM-based QE approaches. Empirical evidence suggests that intermediate Transformer layers in LLMs provide contextual representations that are more aligned with the cross-lingual nature of the QE task. We make resultant models and framework code publicly available for further research, also allowing existing LLM-based MT frameworks to be scaled with QE capabilities.
Authors: Keshav Varadarajan, Tananun Songdechakraiwut
Abstract: Recently, many bias detection methods have been proposed to determine the level of bias a large language model captures. However, tests to identify which parts of a large language model are responsible for bias towards specific groups remain underdeveloped. In this study, we present a method using topological data analysis to identify which heads in GPT-2 contribute to the misrepresentation of identity groups present in the StereoSet dataset. We find that biases for particular categories, such as gender or profession, are concentrated in attention heads that act as hot spots. The metric we propose can also be used to determine which heads capture bias for a specific group within a bias category, and future work could extend this method to help de-bias large language models.
Authors: Joseph T. Colonel, Baihan Lin
Abstract: Word clouds are a common way to summarize qualitative interviews, yet traditional frequency-based methods often fail in conversational contexts: they surface filler words, ignore paraphrase, and fragment semantically related ideas. This limits their usefulness in early-stage analysis, when researchers need fast, interpretable overviews of what participant actually said. We introduce ThemeClouds, an open-source visualization tool that uses large language models (LLMs) to generate thematic, participant-weighted word clouds from dialogue transcripts. The system prompts an LLM to identify concept-level themes across a corpus and then counts how many unique participants mention each topic, yielding a visualization grounded in breadth of mention rather than raw term frequency. Researchers can customize prompts and visualization parameters, providing transparency and control. Using interviews from a user study comparing five recording-device configurations (31 participants; 155 transcripts, Whisper ASR), our approach surfaces more actionable device concerns than frequency clouds and topic-modeling baselines (e.g., LDA, BERTopic). We discuss design trade-offs for integrating LLM assistance into qualitative workflows, implications for interpretability and researcher agency, and opportunities for interactive analyses such as per-condition contrasts (``diff clouds'').
Authors: Jia Deng, Jie Chen, Zhipeng Chen, Daixuan Cheng, Fei Bai, Beichen Zhang, Yinqian Min, Yanzipeng Gao, Wayne Xin Zhao, Ji-Rong Wen
Abstract: Reinforcement learning with verifiable rewards (RLVR) has emerged as a powerful paradigm for enhancing the reasoning capabilities of large language models (LLMs). Unlike traditional RL approaches, RLVR leverages rule-based feedback to guide LLMs in generating and refining complex reasoning chains -- a process critically dependent on effective exploration strategies. While prior work has demonstrated RLVR's empirical success, the fundamental mechanisms governing LLMs' exploration behaviors remain underexplored. This technical report presents a systematic investigation of exploration capacities in RLVR, covering four main aspects: (1) exploration space shaping, where we develop quantitative metrics to characterize LLMs' capability boundaries; (2) entropy-performance exchange, analyzed across training stages, individual instances, and token-level patterns; and (3) RL performance optimization, examining methods to effectively translate exploration gains into measurable improvements. By unifying previously identified insights with new empirical evidence, this work aims to provide a foundational framework for advancing RLVR systems.
Authors: Puspesh Kumar Srivastava, Uddeshya Raj, Praveen Patel, /Shubham Kumar Nigam, Noel Shallum, Arnab Bhattacharya
Abstract: Bail decisions are among the most frequently adjudicated matters in Indian courts, yet they remain plagued by subjectivity, delays, and inconsistencies. With over 75% of India's prison population comprising undertrial prisoners, many from socioeconomically disadvantaged backgrounds, the lack of timely and fair bail adjudication exacerbates human rights concerns and contributes to systemic judicial backlog. In this paper, we present the Indian Bail Prediction System (IBPS), an AI-powered framework designed to assist in bail decision-making by predicting outcomes and generating legally sound rationales based solely on factual case attributes and statutory provisions. We curate and release a large-scale dataset of 150,430 High Court bail judgments, enriched with structured annotations such as age, health, criminal history, crime category, custody duration, statutes, and judicial reasoning. We fine-tune a large language model using parameter-efficient techniques and evaluate its performance across multiple configurations, with and without statutory context, and with RAG. Our results demonstrate that models fine-tuned with statutory knowledge significantly outperform baselines, achieving strong accuracy and explanation quality, and generalize well to a test set independently annotated by legal experts. IBPS offers a transparent, scalable, and reproducible solution to support data-driven legal assistance, reduce bail delays, and promote procedural fairness in the Indian judicial system.
Authors: Ziheng Li, Zhi-Hong Deng
Abstract: Although the LLM-based in-context learning (ICL) paradigm has demonstrated considerable success across various natural language processing tasks, it encounters challenges in event detection. This is because LLMs lack an accurate understanding of event triggers and tend to make over-interpretation, which cannot be effectively corrected through in-context examples alone. In this paper, we focus on the most challenging one-shot setting and propose KeyCP++, a keyword-centric chain-of-thought prompting approach. KeyCP++ addresses the weaknesses of conventional ICL by automatically annotating the logical gaps between input text and detection results for the demonstrations. Specifically, to generate in-depth and meaningful rationale, KeyCP++ constructs a trigger discrimination prompting template. It incorporates the exemplary triggers (a.k.a keywords) into the prompt as the anchor to simply trigger profiling, let LLM propose candidate triggers, and justify each candidate. These propose-and-judge rationales help LLMs mitigate over-reliance on the keywords and promote detection rule learning. Extensive experiments demonstrate the effectiveness of our approach, showcasing significant advancements in one-shot event detection.
Authors: Anirudh Iyengar Kaniyar Narayana Iyengar, Srija Mukhopadhyay, Adnan Qidwai, Shubhankar Singh, Dan Roth, Vivek Gupta
Abstract: We introduce InterChart, a diagnostic benchmark that evaluates how well vision-language models (VLMs) reason across multiple related charts, a task central to real-world applications such as scientific reporting, financial analysis, and public policy dashboards. Unlike prior benchmarks focusing on isolated, visually uniform charts, InterChart challenges models with diverse question types ranging from entity inference and trend correlation to numerical estimation and abstract multi-step reasoning grounded in 2-3 thematically or structurally related charts. We organize the benchmark into three tiers of increasing difficulty: (1) factual reasoning over individual charts, (2) integrative analysis across synthetically aligned chart sets, and (3) semantic inference over visually complex, real-world chart pairs. Our evaluation of state-of-the-art open and closed-source VLMs reveals consistent and steep accuracy declines as chart complexity increases. We find that models perform better when we decompose multi-entity charts into simpler visual units, underscoring their struggles with cross-chart integration. By exposing these systematic limitations, InterChart provides a rigorous framework for advancing multimodal reasoning in complex, multi-visual environments.
Authors: Luyao Zhuang, Qinggang Zhang, Huachi Zhou, Juhua Liu, Qing Li, Xiao Huang
Abstract: Tool learning has emerged as a promising paradigm for large language models (LLMs) to solve many real-world tasks. Nonetheless, with the tool repository rapidly expanding, it is impractical to contain all tools within the limited input length of LLMs. To alleviate these issues, researchers have explored incorporating a tool retrieval module to select the most relevant tools or represent tools as unique tokens within LLM parameters. However, most state-of-the-art methods are under transductive settings, assuming all tools have been observed during training. Such a setting deviates from reality as the real-world tool repository is evolving and incorporates new tools frequently. When dealing with these unseen tools, which refer to tools not encountered during the training phase, these methods are limited by two key issues, including the large distribution shift and the vulnerability of similarity-based retrieval. To this end, inspired by human cognitive processes of mastering unseen tools through discovering and applying the logical information from prior experience, we introduce a novel Logic-Guided Semantic Bridging framework for inductive tool retrieval, namely, LoSemB, which aims to mine and transfer latent logical information for inductive tool retrieval without costly retraining. Specifically, LoSemB contains a logic-based embedding alignment module to mitigate distribution shifts and implements a relational augmented retrieval mechanism to reduce the vulnerability of similarity-based retrieval. Extensive experiments demonstrate that LoSemB achieves advanced performance in inductive settings while maintaining desirable effectiveness in the transductive setting.
Authors: Charlie Wyatt, Aditya Joshi, Flora Salim
Abstract: Transformer-based models primarily rely on Next Token Prediction (NTP), which predicts the next token in a sequence based on the preceding context. However, NTP's focus on single-token prediction often limits a model's ability to plan ahead or maintain long-range coherence, raising questions about how well LLMs can predict longer contexts, such as full sentences within structured documents. While NTP encourages local fluency, it provides no explicit incentive to ensure global coherence across sentence boundaries-an essential skill for reconstructive or discursive tasks. To investigate this, we evaluate three commercial LLMs (GPT-4o, Claude 3.5 Sonnet, and Gemini 2.0 Flash) on Masked Sentence Prediction (MSP) - the task of infilling a randomly removed sentence - from three domains: ROCStories (narrative), Recipe1M (procedural), and Wikipedia (expository). We assess both fidelity (similarity to the original sentence) and cohesiveness (fit within the surrounding context). Our key finding reveals that commercial LLMs, despite their superlative performance in other tasks, are poor at predicting masked sentences in low-structured domains, highlighting a gap in current model capabilities.
Authors: Zhenliang Zhang, Junzhe Zhang, Xinyu Hu, HuiXuan Zhang, Xiaojun Wan
Abstract: Large language models (LLMs) have achieved remarkable success in various tasks, yet they remain vulnerable to faithfulness hallucinations, where the output does not align with the input. In this study, we investigate whether social bias contributes to these hallucinations, a causal relationship that has not been explored. A key challenge is controlling confounders within the context, which complicates the isolation of causality between bias states and hallucinations. To address this, we utilize the Structural Causal Model (SCM) to establish and validate the causality and design bias interventions to control confounders. In addition, we develop the Bias Intervention Dataset (BID), which includes various social biases, enabling precise measurement of causal effects. Experiments on mainstream LLMs reveal that biases are significant causes of faithfulness hallucinations, and the effect of each bias state differs in direction. We further analyze the scope of these causal effects across various models, specifically focusing on unfairness hallucinations, which are primarily targeted by social bias, revealing the subtle yet significant causal effect of bias on hallucination generation.
Authors: Zeyu Yang, Lai Wei, Roman Koshkin, Xi Chen, Satoshi Nakamura
Abstract: This work proposes a grammar-based chunking strategy that segments input streams into semantically complete units by parsing dependency relations (e.g., noun phrase boundaries, verb-object structures) and punctuation features. The method ensures chunk coherence and minimizes semantic fragmentation. Building on this mechanism, we present SASST (Syntax-Aware Simultaneous Speech Translation), an end-to-end framework integrating frozen Whisper encoder and decoder-only LLM. The unified architecture dynamically outputs translation tokens or
Authors: Haoyuan Wu, Haoxing Chen, Xiaodong Chen, Zhanchao Zhou, Tieyuan Chen, Yihong Zhuang, Guoshan Lu, Zenan Huang, Junbo Zhao, Lin Liu, Zhenzhong Lan, Bei Yu, Jianguo Li
Abstract: The Mixture of Experts (MoE) architecture is a cornerstone of modern state-of-the-art (SOTA) large language models (LLMs). MoE models facilitate scalability by enabling sparse parameter activation. However, traditional MoE architecture uses homogeneous experts of a uniform size, activating a fixed number of parameters irrespective of input complexity and thus limiting computational efficiency. To overcome this limitation, we introduce Grove MoE, a novel architecture incorporating experts of varying sizes, inspired by the heterogeneous big.LITTLE CPU architecture. This architecture features novel adjugate experts with a dynamic activation mechanism, enabling model capacity expansion while maintaining manageable computational overhead. Building on this architecture, we present GroveMoE-Base and GroveMoE-Inst, 33B-parameter LLMs developed by applying an upcycling strategy to the Qwen3-30B-A3B-Base model during mid-training and post-training. GroveMoE models dynamically activate 3.14-3.28B parameters based on token complexity and achieve performance comparable to SOTA open-source models of similar or even larger size.
Authors: Yerin Hwang, Dongryeol Lee, Taegwan Kang, Yongil Kim, Kyomin Jung
Abstract: As large language models take on growing roles as automated evaluators in practical settings, a critical question arises: Can individuals persuade an LLM judge to assign unfairly high scores? This study is the first to reveal that strategically embedded persuasive language can bias LLM judges when scoring mathematical reasoning tasks, where correctness should be independent of stylistic variation. Grounded in Aristotle's rhetorical principles, we formalize seven persuasion techniques (Majority, Consistency, Flattery, Reciprocity, Pity, Authority, Identity) and embed them into otherwise identical responses. Across six math benchmarks, we find that persuasive language leads LLM judges to assign inflated scores to incorrect solutions, by up to 8% on average, with Consistency causing the most severe distortion. Notably, increasing model size does not substantially mitigate this vulnerability. Further analysis demonstrates that combining multiple persuasion techniques amplifies the bias, and pairwise evaluation is likewise susceptible. Moreover, the persuasive effect persists under counter prompting strategies, highlighting a critical vulnerability in LLM-as-a-Judge pipelines and underscoring the need for robust defenses against persuasion-based attacks.
Authors: Olga Kellert, Muhammad Imran, Nicholas Hill Matlis, Mahmud Uz Zaman, Carlos G\'omez-Rodr\'iguez
Abstract: This paper summarizes the results of evaluating a compositional approach for Focus Analysis (FA) in Linguistics and Sentiment Analysis (SA) in Natural Language Processing (NLP). While quantitative evaluations of compositional and non-compositional approaches in SA exist in NLP, similar quantitative evaluations are very rare in FA in Linguistics that deal with linguistic expressions representing focus or emphasis such as "it was John who left". We fill this gap in research by arguing that compositional rules in SA also apply to FA because FA and SA are closely related meaning that SA is part of FA. Our compositional approach in SA exploits basic syntactic rules such as rules of modification, coordination, and negation represented in the formalism of Universal Dependencies (UDs) in English and applied to words representing sentiments from sentiment dictionaries. Some of the advantages of our compositional analysis method for SA in contrast to non-compositional analysis methods are interpretability and explainability. We test the accuracy of our compositional approach and compare it with a non-compositional approach VADER that uses simple heuristic rules to deal with negation, coordination and modification. In contrast to previous related work that evaluates compositionality in SA on long reviews, this study uses more appropriate datasets to evaluate compositionality. In addition, we generalize the results of compositional approaches in SA to compositional approaches in FA.
Authors: Yu-Min Tseng, Wei-Lin Chen, Chung-Chi Chen, Hsin-Hsi Chen
Abstract: Textual data annotation, the process of labeling or tagging text with relevant information, is typically costly, time-consuming, and labor-intensive. While large language models (LLMs) have demonstrated their potential as direct alternatives to human annotators for general domains natural language processing (NLP) tasks, their effectiveness on annotation tasks in domains requiring expert knowledge remains underexplored. In this paper, we investigate: whether top-performing LLMs, which might be perceived as having expert-level proficiency in academic and professional benchmarks, can serve as direct alternatives to human expert annotators? To this end, we evaluate both individual LLMs and multi-agent approaches across three highly specialized domains: finance, biomedicine, and law. Specifically, we propose a multi-agent discussion framework to simulate a group of human annotators, where LLMs are tasked to engage in discussions by considering others' annotations and justifications before finalizing their labels. Additionally, we incorporate reasoning models (e.g., o3-mini) to enable a more comprehensive comparison. Our empirical results reveal that: (1) Individual LLMs equipped with inference-time techniques (e.g., chain-of-thought (CoT), self-consistency) show only marginal or even negative performance gains, contrary to prior literature suggesting their broad effectiveness. (2) Overall, reasoning models do not demonstrate statistically significant improvements over non-reasoning models in most settings. This suggests that extended long CoT provides relatively limited benefits for data annotation in specialized domains. (3) Certain model behaviors emerge in the multi-agent discussion environment. For instance, Claude 3.7 Sonnet with thinking rarely changes its initial annotations, even when other agents provide correct annotations or valid reasoning.
Authors: Amrita Singh, H. Suhan Karaca, Aditya Joshi, Hye-young Paik, Jiaojiao Jiang
Abstract: Despite advances in legal NLP, no comprehensive evaluation covering multiple legal-specific LLMs currently exists for contract classification tasks in contract understanding. To address this gap, we present an evaluation of 10 legal-specific LLMs on three English language contract understanding tasks and compare them with 7 general-purpose LLMs. The results show that legal-specific LLMs consistently outperform general-purpose models, especially on tasks requiring nuanced legal understanding. Legal-BERT and Contracts-BERT establish new SOTAs on two of the three tasks, despite having 69% fewer parameters than the best-performing general-purpose LLM. We also identify CaseLaw-BERT and LexLM as strong additional baselines for contract understanding. Our results provide a holistic evaluation of legal-specific LLMs and will facilitate the development of more accurate contract understanding systems.
Authors: Jakub \v{S}m\'id, Pavel P\v{r}ib\'a\v{n}, Pavel Kr\'al
Abstract: Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment analysis task that aims to identify sentiment toward specific aspects of an entity. While large language models (LLMs) have shown strong performance in various natural language processing (NLP) tasks, their capabilities for Czech ABSA remain largely unexplored. In this work, we conduct a comprehensive evaluation of 19 LLMs of varying sizes and architectures on Czech ABSA, comparing their performance in zero-shot, few-shot, and fine-tuning scenarios. Our results show that small domain-specific models fine-tuned for ABSA outperform general-purpose LLMs in zero-shot and few-shot settings, while fine-tuned LLMs achieve state-of-the-art results. We analyze how factors such as multilingualism, model size, and recency influence performance and present an error analysis highlighting key challenges, particularly in aspect term prediction. Our findings provide insights into the suitability of LLMs for Czech ABSA and offer guidance for future research in this area.
Authors: Jakub \v{S}m\'id, Pavel P\v{r}ib\'a\v{n}, Pavel Kr\'al
Abstract: Aspect-based sentiment analysis (ABSA) has received substantial attention in English, yet challenges remain for low-resource languages due to the scarcity of labelled data. Current cross-lingual ABSA approaches often rely on external translation tools and overlook the potential benefits of incorporating a small number of target language examples into training. In this paper, we evaluate the effect of adding few-shot target language examples to the training set across four ABSA tasks, six target languages, and two sequence-to-sequence models. We show that adding as few as ten target language examples significantly improves performance over zero-shot settings and achieves a similar effect to constrained decoding in reducing prediction errors. Furthermore, we demonstrate that combining 1,000 target language examples with English data can even surpass monolingual baselines. These findings offer practical insights for improving cross-lingual ABSA in low-resource and domain-specific settings, as obtaining ten high-quality annotated examples is both feasible and highly effective.
Authors: Chen Cecilia Liu, Hiba Arnaout, Nils Kova\v{c}i\'c, Dana Atzil-Slonim, Iryna Gurevych
Abstract: Large language models (LLMs) show promise in offering emotional support and generating empathetic responses for individuals in distress, but their ability to deliver culturally sensitive support remains underexplored due to lack of resources. In this work, we introduce CultureCare, the first dataset designed for this task, spanning four cultures and including 1729 distress messages, 1523 cultural signals, and 1041 support strategies with fine-grained emotional and cultural annotations. Leveraging CultureCare, we (i) develop and test four adaptation strategies for guiding three state-of-the-art LLMs toward culturally sensitive responses; (ii) conduct comprehensive evaluations using LLM judges, in-culture human annotators, and clinical psychologists; (iii) show that adapted LLMs outperform anonymous online peer responses, and that simple cultural role-play is insufficient for cultural sensitivity; and (iv) explore the application of LLMs in clinical training, where experts highlight their potential in fostering cultural competence in future therapists.
Authors: John C. McDonald, Rosalee Wolfe, Fabrizio Nunnari
Abstract: Non-manual signals in sign languages continue to be a challenge for signing avatars. More specifically, emotional content has been difficult to incorporate because of a lack of a standard method of specifying the avatar's emotional state. This paper explores the application of an intuitive two-parameter representation for emotive non-manual signals to the Paula signing avatar that shows promise for facilitating the linguistic specification of emotional facial expressions in a more coherent manner than previous methods. Users can apply these parameters to control Paula's emotional expressions through a textual representation called the EASIER notation. The representation can allow avatars to express more nuanced emotional states using two numerical parameters. It also has the potential to enable more consistent specification of emotional non-manual signals in linguistic annotations which drive signing avatars.
Authors: Furkan \c{S}ahinu\c{c}, Subhabrata Dutta, Iryna Gurevych
Abstract: Expert domain writing, such as scientific writing, typically demands extensive domain knowledge. Recent advances in LLMs show promising potential in reducing the expert workload. However, evaluating the quality of automatically generated scientific writing is a crucial open issue, as it requires knowledge of domain-specific evaluation criteria and the ability to discern expert preferences. Conventional automatic metrics and LLM-as-a-judge systems are insufficient to grasp expert preferences and domain-specific quality standards. To address this gap and support human-AI collaborative writing, we focus on related work generation, one of the most challenging scientific tasks, as an exemplar. We propose GREP, a multi-turn evaluation framework that integrates classical related work evaluation criteria with expert-specific preferences. Instead of assigning a single score, our framework decomposes the evaluation into fine-grained dimensions. This localized evaluation approach is further augmented with contrastive few-shot examples to provide detailed contextual guidance for the evaluation dimensions. The design principles allow our framework to deliver cardinal assessment of quality, which can facilitate better post-training compared to ordinal preference data. For better accessibility, we design two variants of GREP: a more precise variant with proprietary LLMs as evaluators, and a cheaper alternative with open-weight LLMs. Empirical investigation reveals that our framework is able to assess the quality of related work sections in a much more robust manner compared to standard LLM judges, reflects natural scenarios of scientific writing, and bears a strong correlation with the human expert assessment. We also observe that generations from state-of-the-art LLMs struggle to satisfy validation constraints of a suitable related work section. They (mostly) fail to improve based on feedback as well.
Authors: Changhao Song, Yazhou Zhang, Hui Gao, Ben Yao, Peng Zhang
Abstract: Subjective language understanding refers to a broad set of natural language processing tasks where the goal is to interpret or generate content that conveys personal feelings, opinions, or figurative meanings rather than objective facts. With the advent of large language models (LLMs) such as ChatGPT, LLaMA, and others, there has been a paradigm shift in how we approach these inherently nuanced tasks. In this survey, we provide a comprehensive review of recent advances in applying LLMs to subjective language tasks, including sentiment analysis, emotion recognition, sarcasm detection, humor understanding, stance detection, metaphor interpretation, intent detection, and aesthetics assessment. We begin by clarifying the definition of subjective language from linguistic and cognitive perspectives, and we outline the unique challenges posed by subjective language (e.g. ambiguity, figurativeness, context dependence). We then survey the evolution of LLM architectures and techniques that particularly benefit subjectivity tasks, highlighting why LLMs are well-suited to model subtle human-like judgments. For each of the eight tasks, we summarize task definitions, key datasets, state-of-the-art LLM-based methods, and remaining challenges. We provide comparative insights, discussing commonalities and differences among tasks and how multi-task LLM approaches might yield unified models of subjectivity. Finally, we identify open issues such as data limitations, model bias, and ethical considerations, and suggest future research directions. We hope this survey will serve as a valuable resource for researchers and practitioners interested in the intersection of affective computing, figurative language processing, and large-scale language models.
Authors: Matthias Sperber, Maureen de Seyssel, Jiajun Bao, Matthias Paulik
Abstract: Current speech translation systems, while having achieved impressive accuracies, are rather static in their behavior and do not adapt to real-world situations in ways human interpreters do. In order to improve their practical usefulness and enable interpreting-like experiences, a precise understanding of the nature of human interpreting is crucial. To this end, we discuss human interpreting literature from the perspective of the machine translation field, while considering both operational and qualitative aspects. We identify implications for the development of speech translation systems and argue that there is great potential to adopt many human interpreting principles using recent modeling techniques. We hope that our findings provide inspiration for closing the perceived usability gap, and can motivate progress toward true machine interpreting.
Authors: David Arps, Hassan Sajjad, Laura Kallmeyer
Abstract: Structure-inducing Language Models (SiLM) are trained on a self-supervised language modeling task, and induce a hierarchical sentence representation as a byproduct when processing an input. A wide variety of SiLMs have been proposed. However, these have typically been evaluated on a relatively small scale, and evaluation of these models has systematic gaps and lacks comparability. In this work, we study three different SiLM architectures using both natural language (English) corpora and synthetic bracketing expressions: Structformer (Shen et al., 2021), UDGN (Shen et al., 2022) and GPST (Hu et al., 2024). We compare them with respect to (i) properties of the induced syntactic representations (ii) performance on grammaticality judgment tasks, and (iii) training dynamics. We find that none of the three architectures dominates across all evaluation metrics. However, there are significant differences, in particular with respect to the induced syntactic representations. The Generative Pretrained Structured Transformer (GPST; Hu et al. 2024) performs most consistently across evaluation settings, and outperforms the other models on long-distance dependencies in bracketing expressions. Furthermore, our study shows that small models trained on large amounts of synthetic data provide a useful testbed for evaluating basic model properties.
Authors: Jiaxuan Gao, Wei Fu, Minyang Xie, Shusheng Xu, Chuyi He, Zhiyu Mei, Banghua Zhu, Yi Wu
Abstract: Recent advancements in LLM-based agents have demonstrated remarkable capabilities in handling complex, knowledge-intensive tasks by integrating external tools. Among diverse choices of tools, search tools play a pivotal role in accessing vast external knowledge. However, open-source agents still fall short of achieving expert-level Search Intelligence, the ability to resolve ambiguous queries, generate precise searches, analyze results, and conduct thorough exploration. Existing approaches fall short in scalability, efficiency, and data quality. For example, small turn limits in existing online RL methods, e.g. <=10, restrict complex strategy learning. This paper introduces ASearcher, an open-source project for large-scale RL training of search agents. Our key contributions include: (1) Scalable fully asynchronous RL training that enables long-horizon search while maintaining high training efficiency. (2) A prompt-based LLM agent that autonomously synthesizes high-quality and challenging QAs, creating a large-scale QA dataset. Through RL training, our prompt-based QwQ-32B agent achieves substantial improvements, with 46.7% and 20.8% Avg@4 gains on xBench and GAIA, respectively. Notably, our agent exhibits extreme long-horizon search, with tool calls exceeding 40 turns and output tokens exceeding 150k during training time. With a simple agent design and no external LLMs, ASearcher-Web-QwQ achieves Avg@4 scores of 42.1 on xBench and 52.8 on GAIA, surpassing existing open-source 32B agents. We open-source our models, training data, and codes in https://github.com/inclusionAI/ASearcher.
Authors: Anna Sofia Lippolis, Andrea Giovanni Nuzzolese, Aldo Gangemi
Abstract: Metaphor is a fundamental cognitive mechanism that shapes scientific understanding, enabling the communication of complex concepts while potentially constraining paradigmatic thinking. Despite the prevalence of figurative language in scientific discourse, existing metaphor detection resources primarily focus on general-domain text, leaving a critical gap for domain-specific applications. In this paper, we present the Medical Metaphors Corpus (MCC), a comprehensive dataset of 792 annotated scientific conceptual metaphors spanning medical and biological domains. MCC aggregates metaphorical expressions from diverse sources including peer-reviewed literature, news media, social media discourse, and crowdsourced contributions, providing both binary and graded metaphoricity judgments validated through human annotation. Each instance includes source-target conceptual mappings and perceived metaphoricity scores on a 0-7 scale, establishing the first annotated resource for computational scientific metaphor research. Our evaluation demonstrates that state-of-the-art language models achieve modest performance on scientific metaphor detection, revealing substantial room for improvement in domain-specific figurative language understanding. MCC enables multiple research applications including metaphor detection benchmarking, quality-aware generation systems, and patient-centered communication tools.
Authors: Ryan Wong, Jiawei Wang, Junjie Zhao, Li Chen, Yan Gao, Long Zhang, Xuan Zhou, Zuo Wang, Kai Xiang, Ge Zhang, Wenhao Huang, Yang Wang, Ke Wang
Abstract: From professional research to everyday planning, many tasks are bottlenecked by wide-scale information seeking, which is more repetitive than cognitively complex. With the rapid development of Large Language Models (LLMs), automated search agents powered by LLMs offer a promising solution to liberate humans from this tedious work. However, the capability of these agents to perform such "wide-context" collection reliably and completely remains largely unevaluated due to a lack of suitable benchmarks. To bridge this gap, we introduce WideSearch, a new benchmark engineered to evaluate agent reliability on these large-scale collection tasks. The benchmark features 200 manually curated questions (100 in English, 100 in Chinese) from over 15 diverse domains, grounded in real user queries. Each task requires agents to collect large-scale atomic information, which could be verified one by one objectively, and arrange it into a well-organized output. A rigorous five-stage quality control pipeline ensures the difficulty, completeness, and verifiability of the dataset. We benchmark over 10 state-of-the-art agentic search systems, including single-agent, multi-agent frameworks, and end-to-end commercial systems. Most systems achieve overall success rates near 0\%, with the best performer reaching just 5\%. However, given sufficient time, cross-validation by multiple human testers can achieve a near 100\% success rate. These results demonstrate that present search agents have critical deficiencies in large-scale information seeking, underscoring urgent areas for future research and development in agentic search. Our dataset, evaluation pipeline, and benchmark results have been publicly released at https://widesearch-seed.github.io/
Authors: Mingzi Cao, Xi Wang, Nikolaos Aletras
Abstract: Scaling Large Language Models (LLMs) yields performance gains but incurs substantial training costs. Depth up-scaling offers training efficiency by adding new layers to pre-trained models. However, most existing methods copy or average weights from base layers, neglecting neuron permutation differences. This limitation can potentially cause misalignment that harms performance. Inspired by applying Optimal Transport (OT) for neuron alignment, we propose Optimal Transport Depth Up-Scaling (OpT-DeUS). OpT-DeUS aligns and fuses Transformer blocks in adjacent base layers via OT for new layer creation, to mitigate neuron permutation mismatch between layers. OpT-DeUS achieves better overall performance and offers improved training efficiency than existing methods for continual pre-training and supervised fine-tuning across different model sizes. To further evaluate the impact of interpolation positions, our extensive analysis shows that inserting new layers closer to the top results in higher training efficiency due to shorter back-propagation time while obtaining additional performance gains.
Authors: Fabrizio Nunnari, Cristina Luna Jim\'enez, Rosalee Wolfe, John C. McDonald, Michael Filhol, Eleni Efthimiou, Evita Fotinea, Thomas Hanke
Abstract: The Sign Language Translation and Avatar Technology (SLTAT) workshops continue a series of gatherings to share recent advances in improving deaf / human communication through non-invasive means. This 2025 edition, the 9th since its first appearance in 2011, is hosted by the International Conference on Intelligent Virtual Agents (IVA), giving the opportunity for contamination between two research communities, using digital humans as either virtual interpreters or as interactive conversational agents. As presented in this summary paper, SLTAT sees contributions beyond avatar technologies, with a consistent number of submissions on sign language recognition, and other work on data collection, data analysis, tools, ethics, usability, and affective computing.
Authors: Chun Wang, Chenyang Liu, Wenze Xu, Weihong Deng
Abstract: Conversational systems relying on text-based large language models (LLMs) often overlook paralinguistic cues, essential for understanding emotions and intentions. Speech-language models (SLMs), which use speech as input, are emerging as a promising solution. However, SLMs built by extending frozen LLMs struggle to capture paralinguistic information and exhibit reduced context understanding. We identify entangled information and improper training strategies as key issues. To address these issues, we propose two heterogeneous adapters and suggest a weakly supervised training strategy. Our approach disentangles paralinguistic and linguistic information, enabling SLMs to interpret speech through structured representations. It also preserves contextual understanding by avoiding the generation of task-specific vectors through controlled randomness. This approach trains only the adapters on common datasets, ensuring parameter and data efficiency. Experiments demonstrate competitive performance in emotional conversation tasks, showcasing the model's ability to effectively integrate both paralinguistic and linguistic information within contextual settings.
Authors: Lukas Gehring, Benjamin Paa{\ss}en
Abstract: Recent advancements in Large Language Models (LLMs) and their increased accessibility have made it easier than ever for students to automatically generate texts, posing new challenges for educational institutions. To enforce norms of academic integrity and ensure students' learning, learning analytics methods to automatically detect LLM-generated text appear increasingly appealing. This paper benchmarks the performance of different state-of-the-art detectors in educational contexts, introducing a novel dataset, called Generative Essay Detection in Education (GEDE), containing over 900 student-written essays and over 12,500 LLM-generated essays from various domains. To capture the diversity of LLM usage practices in generating text, we propose the concept of contribution levels, representing students' contribution to a given assignment. These levels range from purely human-written texts, to slightly LLM-improved versions, to fully LLM-generated texts, and finally to active attacks on the detector by "humanizing" generated texts. We show that most detectors struggle to accurately classify texts of intermediate student contribution levels, like LLM-improved human-written texts. Detectors are particularly likely to produce false positives, which is problematic in educational settings where false suspicions can severely impact students' lives. Our dataset, code, and additional supplementary materials are publicly available at https://github.com/lukasgehring/Assessing-LLM-Text-Detection-in-Educational-Contexts.
URLs: https://github.com/lukasgehring/Assessing-LLM-Text-Detection-in-Educational-Contexts.
Authors: Robin Huo, Ewan Dunbar
Abstract: Self-supervised models for speech representation learning now see widespread use for their versatility and performance on downstream tasks, but the effect of model architecture on the linguistic information learned in their representations remains under-studied. This study investigates two such models, HuBERT and wav2vec 2.0, and minimally compares two of their architectural differences: training objective and iterative pseudo-label refinement through multiple training iterations. We find that differences in canonical correlation of hidden representations to word identity, phoneme identity, and speaker identity are explained by training iteration, not training objective. We suggest that future work investigate the reason for the effectiveness of iterative refinement in encoding linguistic information in self-supervised speech representations.
Authors: Jakub \v{S}m\'id, Pavel P\v{r}ib\'a\v{n}, Ond\v{r}ej Pra\v{z}\'ak, Pavel Kr\'al
Abstract: In this paper, we introduce a novel Czech dataset for aspect-based sentiment analysis (ABSA), which consists of 3.1K manually annotated reviews from the restaurant domain. The dataset is built upon the older Czech dataset, which contained only separate labels for the basic ABSA tasks such as aspect term extraction or aspect polarity detection. Unlike its predecessor, our new dataset is specifically designed for more complex tasks, e.g. target-aspect-category detection. These advanced tasks require a unified annotation format, seamlessly linking sentiment elements (labels) together. Our dataset follows the format of the well-known SemEval-2016 datasets. This design choice allows effortless application and evaluation in cross-lingual scenarios, ultimately fostering cross-language comparisons with equivalent counterpart datasets in other languages. The annotation process engaged two trained annotators, yielding an impressive inter-annotator agreement rate of approximately 90%. Additionally, we provide 24M reviews without annotations suitable for unsupervised learning. We present robust monolingual baseline results achieved with various Transformer-based models and insightful error analysis to supplement our contributions. Our code and dataset are freely available for non-commercial research purposes.
Authors: Wenze Xu, Chun Wang, Jiazhen Yu, Sheng Chen, Liang Gao, Weihong Deng
Abstract: Spoken Language Models (SLMs), which extend Large Language Models (LLMs) to perceive speech inputs, have gained increasing attention for their potential to advance speech understanding tasks. However, despite recent progress, studies show that SLMs often struggle to generalize across datasets, even for trained languages and tasks, raising concerns about whether they process speech in a text-like manner as intended. A key challenge underlying this limitation is the modality gap between speech and text representations. The high variability in speech embeddings may allow SLMs to achieve strong in-domain performance by exploiting unintended speech variations, ultimately hindering generalization. To mitigate this modality gap, we introduce Optimal Transport Regularization (OTReg), a method that formulates speech-text alignment as an optimal transport problem and derives a regularization loss to improve SLM training. In each training iteration, OTReg first establishes a structured correspondence between speech and transcript embeddings by determining the optimal transport plan, then incorporates the regularization loss based on this transport plan to optimize SLMs in generating speech embeddings that align more effectively with transcript embeddings. OTReg is lightweight, requiring no additional labels or learnable parameters, and integrates seamlessly into existing SLM training procedures. Extensive multilingual ASR experiments demonstrate that OTReg enhances speech-text alignment, mitigates the modality gap, and consequently improves SLM generalization across diverse datasets.
Authors: Tianyi Zhou, Johanne Medina, Sanjay Chawla
Abstract: Large Language Models (LLMs) are prone to generating fluent but incorrect content, known as confabulation, which poses increasing risks in multi-turn or agentic applications where outputs may be reused as context. In this work, we investigate how in-context information influences model behavior and whether LLMs can identify their unreliable responses. We propose a reliability estimation that leverages token-level uncertainty to guide the aggregation of internal model representations. Specifically, we compute aleatoric and epistemic uncertainty from output logits to identify salient tokens and aggregate their hidden states into compact representations for response-level reliability prediction. Through controlled experiments on open QA benchmarks, we find that correct in-context information improves both answer accuracy and model confidence, while misleading context often induces confidently incorrect responses, revealing a misalignment between uncertainty and correctness. Our probing-based method captures these shifts in model behavior and improves the detection of unreliable outputs across multiple open-source LLMs. These results underscore the limitations of direct uncertainty signals and highlight the potential of uncertainty-guided probing for reliability-aware generation.
Authors: Jun Wang, Zaifu Zhan, Qixin Zhang, Mingquan Lin, Meijia Song, Rui Zhang
Abstract: Recent progress in large language models (LLMs) has leveraged their in-context learning (ICL) abilities to enable quick adaptation to unseen biomedical NLP tasks. By incorporating only a few input-output examples into prompts, LLMs can rapidly perform these new tasks. While the impact of these demonstrations on LLM performance has been extensively studied, most existing approaches prioritize representativeness over diversity when selecting examples from large corpora. To address this gap, we propose Dual-Div, a diversity-enhanced data-efficient framework for demonstration selection in biomedical ICL. Dual-Div employs a two-stage retrieval and ranking process: First, it identifies a limited set of candidate examples from a corpus by optimizing both representativeness and diversity (with optional annotation for unlabeled data). Second, it ranks these candidates against test queries to select the most relevant and non-redundant demonstrations. Evaluated on three biomedical NLP tasks (named entity recognition (NER), relation extraction (RE), and text classification (TC)) using LLaMA 3.1 and Qwen 2.5 for inference, along with three retrievers (BGE-Large, BMRetriever, MedCPT), Dual-Div consistently outperforms baselines-achieving up to 5% higher macro-F1 scores-while demonstrating robustness to prompt permutations and class imbalance. Our findings establish that diversity in initial retrieval is more critical than ranking-stage optimization, and limiting demonstrations to 3-5 examples maximizes performance efficiency.
Authors: Wentao Jiang, Xiang Feng, Zengmao Wang, Yong Luo, Pingbo Xu, Zhe Chen, Bo Du, Jing Zhang
Abstract: Reinforcement learning (RL) is emerging as a powerful paradigm for enabling large language models (LLMs) to perform complex reasoning tasks. Recent advances indicate that integrating RL with retrieval-augmented generation (RAG) allows LLMs to dynamically incorporate external knowledge, leading to more informed and robust decision making. However, we identify a critical challenge during policy-driven trajectory sampling: LLMs are frequently trapped in unproductive reasoning paths, which we refer to as "dead ends", committing to overconfident yet incorrect conclusions. This severely hampers exploration and undermines effective policy optimization. To address this challenge, we propose REX-RAG (Reasoning Exploration with Policy Correction in Retrieval-Augmented Generation), a novel framework that explores alternative reasoning paths while maintaining rigorous policy learning through principled distributional corrections. Our approach introduces two key innovations: (1) Mixed Sampling Strategy, which combines a novel probe sampling method with exploratory prompts to escape dead ends; and (2) Policy Correction Mechanism, which employs importance sampling to correct distribution shifts induced by mixed sampling, thereby mitigating gradient estimation bias. We evaluate it on seven question-answering benchmarks, and the experimental results show that REX-RAG achieves average performance gains of 5.1% on Qwen2.5-3B and 3.6% on Qwen2.5-7B over strong baselines, demonstrating competitive results across multiple datasets. The code is publicly available at https://github.com/MiliLab/REX-RAG.
Authors: Mandira Sawkar, Samay U. Shetty, Deepak Pandita, Tharindu Cyril Weerasooriya, Christopher M. Homan
Abstract: The Learning With Disagreements (LeWiDi) 2025 shared task is to model annotator disagreement through soft label distribution prediction and perspectivist evaluation, modeling annotators. We adapt DisCo (Distribution from Context), a neural architecture that jointly models item-level and annotator-level label distributions, and present detailed analysis and improvements. In this paper, we extend the DisCo by incorporating annotator metadata, enhancing input representations, and modifying the loss functions to capture disagreement patterns better. Through extensive experiments, we demonstrate substantial improvements in both soft and perspectivist evaluation metrics across three datasets. We also conduct in-depth error and calibration analyses, highlighting the conditions under which improvements occur. Our findings underscore the value of disagreement-aware modeling and offer insights into how system components interact with the complexity of human-annotated data.
Authors: Bangsheng Tang, Carl Chengyan Fu, Fei Kou, Grigory Sizov, Haoci Zhang, Jason Park, Jiawen Liu, Jie You, Qirui Yang, Sachin Mehta, Shengyong Cai, Xiaodong Wang, Xingyu Liu, Yunlu Li, Yanjun Zhou, Wei Wei, Zhiwei Zhao, Zixi Qi, Adolfo Victoria, Aya Ibrahim, Bram Wasti, Changkyu Kim, Daniel Haziza, Fei Sun, Giancarlo Delfin, Emily Guo, Jialin Ouyang, Jaewon Lee, Jianyu Huang, Jeremy Reizenstein, Lu Fang, Quinn Zhu, Ria Verma, Vlad Mihailescu, Xingwen Guo, Yan Cui, Ye Hu, Yejin Lee
Abstract: Speculative decoding is a standard method for accelerating the inference speed of large language models. However, scaling it for production environments poses several engineering challenges, including efficiently implementing different operations (e.g., tree attention and multi-round speculative decoding) on GPU. In this paper, we detail the training and inference optimization techniques that we have implemented to enable EAGLE-based speculative decoding at a production scale for Llama models. With these changes, we achieve a new state-of-the-art inference latency for Llama models. For example, Llama4 Maverick decodes at a speed of about 4 ms per token (with a batch size of one) on 8 NVIDIA H100 GPUs, which is 10% faster than the previously best known method. Furthermore, for EAGLE-based speculative decoding, our optimizations enable us to achieve a speed-up for large batch sizes between 1.4x and 2.0x at production scale.
Authors: Kyle Moore, Jesse Roberts, Daryl Watson
Abstract: There has been much recent interest in evaluating large language models for uncertainty calibration to facilitate model control and modulate user trust. Inference time uncertainty, which may provide a real-time signal to the model or external control modules, is particularly important for applying these concepts to improve LLM-user experience in practice. While many of the existing papers consider model calibration, comparatively little work has sought to evaluate how closely model uncertainty aligns to human uncertainty. In this work, we evaluate a collection of inference-time uncertainty measures, using both established metrics and novel variations, to determine how closely they align with both human group-level uncertainty and traditional notions of model calibration. We find that numerous measures show evidence of strong alignment to human uncertainty, even despite the lack of alignment to human answer preference. For those successful metrics, we find moderate to strong evidence of model calibration in terms of both correctness correlation and distributional analysis.
Authors: Zhuohao Yu, Xingru Jiang, Weizheng Gu, Yidong Wang, Shikun Zhang, Wei Ye
Abstract: Watermarking LLM-generated text is critical for content attribution and misinformation prevention. However, existing methods compromise text quality, require white-box model access and logit manipulation. These limitations exclude API-based models and multilingual scenarios. We propose SAEMark, a general framework for post-hoc multi-bit watermarking that embeds personalized messages solely via inference-time, feature-based rejection sampling without altering model logits or requiring training. Our approach operates on deterministic features extracted from generated text, selecting outputs whose feature statistics align with key-derived targets. This framework naturally generalizes across languages and domains while preserving text quality through sampling LLM outputs instead of modifying. We provide theoretical guarantees relating watermark success probability and compute budget that hold for any suitable feature extractor. Empirically, we demonstrate the framework's effectiveness using Sparse Autoencoders (SAEs), achieving superior detection accuracy and text quality. Experiments across 4 datasets show SAEMark's consistent performance, with 99.7% F1 on English and strong multi-bit detection accuracy. SAEMark establishes a new paradigm for scalable watermarking that works out-of-the-box with closed-source LLMs while enabling content attribution.
Authors: Shansong Wang, Mingzhe Hu, Qiang Li, Mojtaba Safari, Xiaofeng Yang
Abstract: Recent advances in large language models (LLMs) have enabled general-purpose systems to perform increasingly complex domain-specific reasoning without extensive fine-tuning. In the medical domain, decision-making often requires integrating heterogeneous information sources, including patient narratives, structured data, and medical images. This study positions GPT-5 as a generalist multimodal reasoner for medical decision support and systematically evaluates its zero-shot chain-of-thought reasoning performance on both text-based question answering and visual question answering tasks under a unified protocol. We benchmark GPT-5, GPT-5-mini, GPT-5-nano, and GPT-4o-2024-11-20 against standardized splits of MedQA, MedXpertQA (text and multimodal), MMLU medical subsets, USMLE self-assessment exams, and VQA-RAD. Results show that GPT-5 consistently outperforms all baselines, achieving state-of-the-art accuracy across all QA benchmarks and delivering substantial gains in multimodal reasoning. On MedXpertQA MM, GPT-5 improves reasoning and understanding scores by +29.62% and +36.18% over GPT-4o, respectively, and surpasses pre-licensed human experts by +24.23% in reasoning and +29.40% in understanding. In contrast, GPT-4o remains below human expert performance in most dimensions. A representative case study demonstrates GPT-5's ability to integrate visual and textual cues into a coherent diagnostic reasoning chain, recommending appropriate high-stakes interventions. Our results show that, on these controlled multimodal reasoning benchmarks, GPT-5 moves from human-comparable to above human-expert performance. This improvement may substantially inform the design of future clinical decision-support systems.
Authors: Yunna Cai, Fan Wang, Haowei Wang, Kun Wang, Kailai Yang, Sophia Ananiadou, Moyan Li, Mingming Fan
Abstract: Evaluating the safety alignment of LLM responses in high-risk mental health dialogues is particularly difficult due to missing gold-standard answers and the ethically sensitive nature of these interactions. To address this challenge, we propose PsyCrisis-Bench, a reference-free evaluation benchmark based on real-world Chinese mental health dialogues. It evaluates whether the model responses align with the safety principles defined by experts. Specifically designed for settings without standard references, our method adopts a prompt-based LLM-as-Judge approach that conducts in-context evaluation using expert-defined reasoning chains grounded in psychological intervention principles. We employ binary point-wise scoring across multiple safety dimensions to enhance the explainability and traceability of the evaluation. Additionally, we present a manually curated, high-quality Chinese-language dataset covering self-harm, suicidal ideation, and existential distress, derived from real-world online discourse. Experiments on 3600 judgments show that our method achieves the highest agreement with expert assessments and produces more interpretable evaluation rationales compared to existing approaches. Our dataset and evaluation tool are publicly available to facilitate further research.
Authors: Jiahao Zhao, Liwei Dong
Abstract: Unlimited, or so-called helpful-only language models are trained without safety alignment constraints and never refuse user queries. They are widely used by leading AI companies as internal tools for red teaming and alignment evaluation. For example, if a safety-aligned model produces harmful outputs similar to an unlimited model, this indicates alignment failures that require further attention. Despite their essential role in assessing alignment, such models are not available to the research community. We introduce Jinx, a helpful-only variant of popular open-weight LLMs. Jinx responds to all queries without refusals or safety filtering, while preserving the base model's capabilities in reasoning and instruction following. It provides researchers with an accessible tool for probing alignment failures, evaluating safety boundaries, and systematically studying failure modes in language model safety.
Authors: Rachel K. Luu, Jingyu Deng, Mohammed Shahrudin Ibrahim, Nam-Joon Cho, Ming Dao, Subra Suresh, Markus J. Buehler
Abstract: Large language models (LLMs) have reshaped the research landscape by enabling new approaches to knowledge retrieval and creative ideation. Yet their application in discipline-specific experimental science, particularly in highly multi-disciplinary domains like materials science, remains limited. We present a first-of-its-kind framework that integrates generative AI with literature from hitherto-unconnected fields such as plant science, biomimetics, and materials engineering to extract insights and design experiments for materials. We focus on humidity-responsive systems such as pollen-based materials and Rhapis excelsa (broadleaf lady palm) leaves, which exhibit self-actuation and adaptive performance. Using a suite of AI tools, including a fine-tuned model (BioinspiredLLM), Retrieval-Augmented Generation (RAG), agentic systems, and a Hierarchical Sampling strategy, we extract structure-property relationships and translate them into new classes of bioinspired materials. Structured inference protocols generate and evaluate hundreds of hypotheses from a single query, surfacing novel and experimentally tractable ideas. We validate our approach through real-world implementation: LLM-generated procedures, materials designs, and mechanical predictions were tested in the laboratory, culminating in the fabrication of a novel pollen-based adhesive with tunable morphology and measured shear strength, establishing a foundation for future plant-derived adhesive design. This work demonstrates how AI-assisted ideation can drive real-world materials design and enable effective human-AI collaboration.
Authors: Catherine Yeh, Tara Menon, Robin Singh Arya, Helen He, Moira Weigel, Fernanda Vi\'egas, Martin Wattenberg
Abstract: Analyzing literature involves tracking interactions between characters, locations, and themes. Visualization has the potential to facilitate the mapping and analysis of these complex relationships, but capturing structured information from unstructured story data remains a challenge. As large language models (LLMs) continue to advance, we see an opportunity to use their text processing and analysis capabilities to augment and reimagine existing storyline visualization techniques. Toward this goal, we introduce an LLM-driven data parsing pipeline that automatically extracts relevant narrative information from novels and scripts. We then apply this pipeline to create Story Ribbons, an interactive visualization system that helps novice and expert literary analysts explore detailed character and theme trajectories at multiple narrative levels. Through pipeline evaluations and user studies with Story Ribbons on 36 literary works, we demonstrate the potential of LLMs to streamline narrative visualization creation and reveal new insights about familiar stories. We also describe current limitations of AI-based systems, and interaction motifs designed to address these issues.
Authors: Jinsung Yoon, Wooyeol Jeong, Jio Gim, Young-Joo Suh
Abstract: Emotional voice conversion (EVC) aims to modify the emotional style of speech while preserving its linguistic content. In practical EVC, controllability, the ability to independently control speaker identity and emotional style using distinct references, is crucial. However, existing methods often struggle to fully disentangle these attributes and lack the ability to model fine-grained emotional expressions such as temporal dynamics. We propose Maestro-EVC, a controllable EVC framework that enables independent control of content, speaker identity, and emotion by effectively disentangling each attribute from separate references. We further introduce a temporal emotion representation and an explicit prosody modeling with prosody augmentation to robustly capture and transfer the temporal dynamics of the target emotion, even under prosody-mismatched conditions. Experimental results confirm that Maestro-EVC achieves high-quality, controllable, and emotionally expressive speech synthesis.
Authors: Lixuan He, Jie Feng, Yong Li
Abstract: Large Language Models (LLMs) are typically fine-tuned for reasoning tasks through a two-stage pipeline of Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL), a process fraught with catastrophic forgetting and suboptimal trade-offs between imitation and exploration. Recent single-stage methods attempt to unify SFT and RL using heuristics, but lack a principled mechanism for dynamically balancing the two paradigms. In this paper, we reframe this challenge through the theoretical lens of \textbf{implicit rewards}, viewing SFT and RL not as distinct methods but as complementary reward signals. We introduce \textbf{Adaptive Meta Fine-Tuning (AMFT)}, a novel single-stage algorithm that learns the optimal balance between SFT's implicit, path-level reward and RL's explicit, outcome-based reward. The core of AMFT is a \textbf{meta-gradient adaptive weight controller} that treats the SFT-RL balance as a learnable parameter, dynamically optimizing it to maximize long-term task performance. This forward-looking approach, regularized by policy entropy for stability, autonomously discovers an effective training curriculum. We conduct a comprehensive evaluation on challenging benchmarks spanning mathematical reasoning, abstract visual reasoning (General Points), and vision-language navigation (V-IRL). AMFT consistently establishes a new state-of-the-art and demonstrats superior generalization on out-of-distribution (OOD) tasks. Ablation studies and training dynamic analysis confirm that the meta-learning controller is crucial for AMFT's stability, sample efficiency, and performance, offering a more principled and effective paradigm for LLM alignment. Our codes are open-sourced via https://github.com/hlxtsyj/AMFT.
Authors: Keyu Li, Mohan Jiang, Dayuan Fu, Yunze Wu, Xiangkun Hu, Dequan Wang, Pengfei Liu
Abstract: The rapid advancement of large language models has fundamentally shifted the bottleneck in AI development from computational power to data availability-with countless valuable datasets remaining hidden across specialized repositories, research appendices, and domain platforms. As reasoning capabilities and deep research methodologies continue to evolve, a critical question emerges: can AI agents transcend conventional search to systematically discover any dataset that meets specific user requirements, enabling truly autonomous demand-driven data curation? We introduce DatasetResearch, the first comprehensive benchmark evaluating AI agents' ability to discover and synthesize datasets from 208 real-world demands across knowledge-intensive and reasoning-intensive tasks. Our tri-dimensional evaluation framework reveals a stark reality: even advanced deep research systems achieve only 22% score on our challenging DatasetResearch-pro subset, exposing the vast gap between current capabilities and perfect dataset discovery. Our analysis uncovers a fundamental dichotomy-search agents excel at knowledge tasks through retrieval breadth, while synthesis agents dominate reasoning challenges via structured generation-yet both catastrophically fail on "corner cases" outside existing distributions. These findings establish the first rigorous baseline for dataset discovery agents and illuminate the path toward AI systems capable of finding any dataset in the digital universe. Our benchmark and comprehensive analysis provide the foundation for the next generation of self-improving AI systems and are publicly available at https://github.com/GAIR-NLP/DatasetResearch.
Authors: Andrei Andrusenko, Vladimir Bataev, Lilit Grigoryan, Vitaly Lavrukhin, Boris Ginsburg
Abstract: Recognizing specific key phrases is an essential task for contextualized Automatic Speech Recognition (ASR). However, most existing context-biasing approaches have limitations associated with the necessity of additional model training, significantly slow down the decoding process, or constrain the choice of the ASR system type. This paper proposes a universal ASR context-biasing framework that supports all major types: CTC, Transducers, and Attention Encoder-Decoder models. The framework is based on a GPU-accelerated word boosting tree, which enables it to be used in shallow fusion mode for greedy and beam search decoding without noticeable speed degradation, even with a vast number of key phrases (up to 20K items). The obtained results showed high efficiency of the proposed method, surpassing the considered open-source context-biasing approaches in accuracy and decoding speed. Our context-biasing framework is open-sourced as a part of the NeMo toolkit.
Authors: Shengtao Wen, Haodong Chen, Yadong Wang, Zhongying Pan, Xiang Chen, Yu Tian, Bo Qian, Dong Liang, Sheng-Jun Huang
Abstract: Knowledge editing (KE) provides a scalable approach for updating factual knowledge in large language models without full retraining. While previous studies have demonstrated effectiveness in general domains and medical QA tasks, little attention has been paid to KE in multimodal medical scenarios. Unlike text-only settings, medical KE demands integrating updated knowledge with visual reasoning to support safe and interpretable clinical decisions. To address this gap, we propose MultiMedEdit, the first benchmark tailored to evaluating KE in clinical multimodal tasks. Our framework spans both understanding and reasoning task types, defines a three-dimensional metric suite (reliability, generality, and locality), and supports cross-paradigm comparisons across general and domain-specific models. We conduct extensive experiments under single-editing and lifelong-editing settings. Results suggest that current methods struggle with generalization and long-tail reasoning, particularly in complex clinical workflows. We further present an efficiency analysis (e.g., edit latency, memory footprint), revealing practical trade-offs in real-world deployment across KE paradigms. Overall, MultiMedEdit not only reveals the limitations of current approaches but also provides a solid foundation for developing clinically robust knowledge editing techniques in the future.
Authors: Wenhan Liu, Xinyu Ma, Weiwei Sun, Yutao Zhu, Yuchen Li, Dawei Yin, Zhicheng Dou
Abstract: Large Language Model (LLM) based listwise ranking has shown superior performance in many passage ranking tasks. With the development of Large Reasoning Models, many studies have demonstrated that step-by-step reasoning during test-time helps improve listwise ranking performance. However, due to the scarcity of reasoning-intensive training data, existing rerankers perform poorly in many complex ranking scenarios and the ranking ability of reasoning-intensive rerankers remains largely underdeveloped. In this paper, we first propose an automated reasoning-intensive training data synthesis framework, which sources training queries and passages from diverse domains and applies DeepSeek-R1 to generate high-quality training labels. A self-consistency data filtering mechanism is designed to ensure the data quality. To empower the listwise reranker with strong reasoning ability, we further propose a two-stage post-training approach, which includes a cold-start supervised fine-tuning (SFT) stage for reasoning pattern learning and a reinforcement learning (RL) stage for further ranking ability enhancement. During the RL stage, based on the nature of listwise ranking, we design a multi-view ranking reward, which is more effective than a ranking metric-based reward. Extensive experiments demonstrate that our trained reasoning-intensive reranker \textbf{ReasonRank} outperforms existing baselines significantly and also achieves much lower latency than pointwise reranker Rank1. \textbf{Through further experiments, our ReasonRank has achieved state-of-the-art (SOTA) performance 40.6 on the BRIGHT leaderboard\footnote{https://brightbenchmark.github.io/}.} Our codes are available at https://github.com/8421BCD/ReasonRank.
URLs: https://brightbenchmark.github.io/, https://github.com/8421BCD/ReasonRank.
Authors: Mohammadreza Daviran, Brian Lin, Davood Rafiei
Abstract: We introduce SQL-Exchange, a framework for mapping SQL queries across different database schemas by preserving the source query structure while adapting domain-specific elements to align with the target schema. We investigate the conditions under which such mappings are feasible and beneficial, and examine their impact on enhancing the in-context learning performance of text-to-SQL systems as a downstream task. Our comprehensive evaluation across multiple model families and benchmark datasets--assessing structural alignment with source queries, execution validity on target databases, and semantic correctness--demonstrates that SQL-Exchange is effective across a wide range of schemas and query types. Our results further show that using mapped queries as in-context examples consistently improves text-to-SQL performance over using queries from the source schema.
Authors: Chaoqun Cui, Caiyan Jia
Abstract: Rumor detection on social media has become increasingly important. Most existing graph-based models presume rumor propagation trees (RPTs) have deep structures and learn sequential stance features along branches. However, through statistical analysis on real-world datasets, we find RPTs exhibit wide structures, with most nodes being shallow 1-level replies. To focus learning on intensive substructures, we propose Rumor Adaptive Graph Contrastive Learning (RAGCL) method with adaptive view augmentation guided by node centralities. We summarize three principles for RPT augmentation: 1) exempt root nodes, 2) retain deep reply nodes, 3) preserve lower-level nodes in deep sections. We employ node dropping, attribute masking and edge dropping with probabilities from centrality-based importance scores to generate views. A graph contrastive objective then learns robust rumor representations. Extensive experiments on four benchmark datasets demonstrate RAGCL outperforms state-of-the-art methods. Our work reveals the wide-structure nature of RPTs and contributes an effective graph contrastive learning approach tailored for rumor detection through principled adaptive augmentation. The proposed principles and augmentation techniques can potentially benefit other applications involving tree-structured graphs.
Authors: Chaoqun Cui, Caiyan Jia
Abstract: Current rumor detection methods based on propagation structure learning predominately treat rumor detection as a class-balanced classification task on limited labeled data. However, real-world social media data exhibits an imbalanced distribution with a minority of rumors among massive regular posts. To address the data scarcity and imbalance issues, we construct two large-scale conversation datasets from Weibo and Twitter and analyze the domain distributions. We find obvious differences between rumor and non-rumor distributions, with non-rumors mostly in entertainment domains while rumors concentrate in news, indicating the conformity of rumor detection to an anomaly detection paradigm. Correspondingly, we propose the Anomaly Detection framework with Graph Supervised Contrastive Learning (AD-GSCL). It heuristically treats unlabeled data as non-rumors and adapts graph contrastive learning for rumor detection. Extensive experiments demonstrate AD-GSCL's superiority under class-balanced, imbalanced, and few-shot conditions. Our findings provide valuable insights for real-world rumor detection featuring imbalanced data distributions.
Authors: Yi Tang, Kaini Wang, Yang Chen, Guangquan Zhou
Abstract: Developing general artificial intelligence (AI) systems to support endoscopic image diagnosis is an emerging research priority. Existing methods based on large-scale pretraining often lack unified coordination across tasks and struggle to handle the multi-step processes required in complex clinical workflows. While AI agents have shown promise in flexible instruction parsing and tool integration across domains, their potential in endoscopy remains underexplored. To address this gap, we propose EndoAgent, the first memory-guided agent for vision-to-decision endoscopic analysis that integrates iterative reasoning with adaptive tool selection and collaboration. Built on a dual-memory design, it enables sophisticated decision-making by ensuring logical coherence through short-term action tracking and progressively enhancing reasoning acuity through long-term experiential learning. To support diverse clinical tasks, EndoAgent integrates a suite of expert-designed tools within a unified reasoning loop. We further introduce EndoAgentBench, a benchmark of 5,709 visual question-answer pairs that assess visual understanding and language generation capabilities in realistic scenarios. Extensive experiments show that EndoAgent consistently outperforms both general and medical multimodal models, exhibiting its strong flexibility and reasoning capabilities.
Authors: Lilit Grigoryan, Vladimir Bataev, Nikolay Karpov, Andrei Andrusenko, Vitaly Lavrukhin, Boris Ginsburg
Abstract: While beam search improves speech recognition quality over greedy decoding, standard implementations are slow, often sequential, and CPU-bound. To fully leverage modern hardware capabilities, we present a novel open-source FlexCTC toolkit for fully GPU-based beam decoding, designed for Connectionist Temporal Classification (CTC) models. Developed entirely in Python and PyTorch, it offers a fast, user-friendly, and extensible alternative to traditional C++, CUDA, or WFST-based decoders. The toolkit features a high-performance, fully batched GPU implementation with eliminated CPU-GPU synchronization and minimized kernel launch overhead via CUDA Graphs. It also supports advanced contextualization techniques, including GPU-powered N-gram language model fusion and phrase-level boosting. These features enable accurate and efficient decoding, making them suitable for both research and production use.
Authors: Kepu Zhang, Teng Shi, Weijie Yu, Jun Xu
Abstract: Personalized retrieval-augmented generation (RAG) aims to produce user-tailored responses by incorporating retrieved user profiles alongside the input query. Existing methods primarily focus on improving retrieval and rely on large language models (LLMs) to implicitly integrate the retrieved context with the query. However, such models are often sensitive to retrieval quality and may generate responses that are misaligned with user preferences. To address this limitation, we propose PrLM, a reinforcement learning framework that trains LLMs to explicitly reason over retrieved user profiles. Guided by a contrastively trained personalization reward model, PrLM effectively learns from user responses without requiring annotated reasoning paths. Experiments on three personalized text generation datasets show that PrLM outperforms existing methods and remains robust across varying numbers of retrieved profiles and different retrievers.
Authors: Rubing Chen, Jiaxin Wu, Jian Wang, Xulu Zhang, Wenqi Fan, Chenghua Lin, Xiao-Yong Wei, Qing Li
Abstract: Numerous benchmarks have been built to evaluate the domain-specific abilities of large language models (LLMs), highlighting the need for effective and efficient benchmark construction. Existing domain-specific benchmarks primarily focus on the scaling law, relying on massive corpora for supervised fine-tuning or generating extensive question sets for broad coverage. However, the impact of corpus and question-answer (QA) set design on the precision and recall of domain-specific LLMs remains unexplored. In this paper, we address this gap and demonstrate that the scaling law is not always the optimal principle for benchmark construction in specific domains. Instead, we propose Comp-Comp, an iterative benchmarking framework based on a comprehensiveness-compactness principle. Here, comprehensiveness ensures semantic recall of the domain, while compactness enhances precision, guiding both corpus and QA set construction. To validate our framework, we conducted a case study in a well-renowned university, resulting in the creation of XUBench, a large-scale and comprehensive closed-domain benchmark. Although we use the academic domain as the case in this work, our Comp-Comp framework is designed to be extensible beyond academia, providing valuable insights for benchmark construction across various domains.
Authors: Jesse Ponnock
Abstract: Given recent breakthroughs in Generative Artificial Intelligence (GAI) and Large Language Models (LLMs), more and more professional services are being augmented through Artificial Intelligence (AI), which once seemed impossible to automate. This paper presents a modular model for leveraging GAI in developing strategic plans for large scale government organizations and evaluates leading machine learning techniques in their application towards one of the identified modules. Specifically, the performance of BERTopic and Non-negative Matrix Factorization (NMF) are evaluated in their ability to use topic modeling to generate themes representative of Vision Elements within a strategic plan. To accomplish this, BERTopic and NMF models are trained using a large volume of reports from the Government Accountability Office (GAO). The generated topics from each model are then scored for similarity against the Vision Elements of a published strategic plan and the results are compared. Our results show that these techniques are capable of generating themes similar to 100% of the elements being evaluated against. Further, we conclude that BERTopic performs best in this application with more than half of its correlated topics achieving a "medium" or "strong" correlation. A capability of GAI-enabled strategic plan development impacts a multi-billion dollar industry and assists the federal government in overcoming regulatory requirements which are crucial to the public good. Further work will focus on the operationalization of the concept proven in this study as well as viability of the remaining modules in the proposed model for GAI-generated strategic plans.
Authors: Jinyuan Fang, Yanwen Peng, Xi Zhang, Yingxu Wang, Xinhao Yi, Guibin Zhang, Yi Xu, Bin Wu, Siwei Liu, Zihao Li, Zhaochun Ren, Nikos Aletras, Xi Wang, Han Zhou, Zaiqiao Meng
Abstract: Recent advances in large language models have sparked growing interest in AI agents capable of solving complex, real-world tasks. However, most existing agent systems rely on manually crafted configurations that remain static after deployment, limiting their ability to adapt to dynamic and evolving environments. To this end, recent research has explored agent evolution techniques that aim to automatically enhance agent systems based on interaction data and environmental feedback. This emerging direction lays the foundation for self-evolving AI agents, which bridge the static capabilities of foundation models with the continuous adaptability required by lifelong agentic systems. In this survey, we provide a comprehensive review of existing techniques for self-evolving agentic systems. Specifically, we first introduce a unified conceptual framework that abstracts the feedback loop underlying the design of self-evolving agentic systems. The framework highlights four key components: System Inputs, Agent System, Environment, and Optimisers, serving as a foundation for understanding and comparing different strategies. Based on this framework, we systematically review a wide range of self-evolving techniques that target different components of the agent system. We also investigate domain-specific evolution strategies developed for specialised fields such as biomedicine, programming, and finance, where optimisation objectives are tightly coupled with domain constraints. In addition, we provide a dedicated discussion on the evaluation, safety, and ethical considerations for self-evolving agentic systems, which are critical to ensuring their effectiveness and reliability. This survey aims to provide researchers and practitioners with a systematic understanding of self-evolving AI agents, laying the foundation for the development of more adaptive, autonomous, and lifelong agentic systems.
Authors: Yueyi Wang, Qiyao Wei
Abstract: In this study, we wish to showcase the unique utility of large language models (LLMs) in financial semantic annotation and alpha signal discovery. Leveraging a corpus of company-related tweets, we use an LLM to automatically assign multi-label event categories to high-sentiment-intensity tweets. We align these labeled sentiment signals with forward returns over 1-to-7-day horizons to evaluate their statistical efficacy and market tradability. Our experiments reveal that certain event labels consistently yield negative alpha, with Sharpe ratios as low as -0.38 and information coefficients exceeding 0.05, all statistically significant at the 95\% confidence level. This study establishes the feasibility of transforming unstructured social media text into structured, multi-label event variables. A key contribution of this work is its commitment to transparency and reproducibility; all code and methodologies are made publicly available. Our results provide compelling evidence that social media sentiment is a valuable, albeit noisy, signal in financial forecasting and underscore the potential of open-source frameworks to democratize algorithmic trading research.
Authors: Stefan Szeider
Abstract: Translating natural language problem descriptions into formal constraint models remains a fundamental challenge in constraint programming, requiring deep expertise in both the problem domain and modeling frameworks. Previous approaches to automating this translation have employed fixed workflows with predetermined modeling steps, failing on a significant number of benchmark problems. We present a new approach using a pure agentic strategy without any fixed pipeline. We developed a general-purpose Python coding agent based on the ReAct (Reason and Act) principle, utilizing a persistent IPython kernel for stateful code execution and iterative development. Rather than embedding constraint programming logic into the agent architecture, domain-specific expertise is injected solely through a carefully crafted project prompt. The agent combines this prompt-encoded knowledge with access to file operations and code execution tools, enabling it to test hypotheses, debug failures, and verify solutions dynamically. Implemented in just a few hundred lines of code, this architecture successfully solves all 101 problems of the CP-Bench constraint programming benchmark set. The results suggest that constraint modeling tasks require the combination of general coding tools and domain expertise encoded in prompts, rather than specialized agent architectures or predefined workflows.
Authors: Alexander Duffy, Samuel J Paech, Ishana Shastri, Elizabeth Karpinski, Baptiste Alloui-Cros, Tyler Marques, Matthew Lyle Olson
Abstract: We present the first evaluation harness that enables any out-of-the-box, local, Large Language Models (LLMs) to play full-press Diplomacy without fine-tuning or specialized training. Previous work required frontier LLMs, or fine-tuning, due to the high complexity and information density of Diplomacy's game state. Combined with the high variance of matches, these factors made Diplomacy prohibitive for study. In this work, we used data-driven iteration to optimize a textual game state representation such that a 24B model can reliably complete matches without any fine tuning. We develop tooling to facilitate hypothesis testing and statistical analysis, and we present case studies on persuasion, aggressive playstyles, and performance across a range of models. We conduct a variety of experiments across many popular LLMs, finding the larger models perform the best, but the smaller models still play adequately. We also introduce Critical State Analysis: an experimental protocol for rapidly iterating and analyzing key moments in a game at depth. Our harness democratizes the evaluation of strategic reasoning in LLMs by eliminating the need for fine-tuning, and it provides insights into how these capabilities emerge naturally from widely used LLMs. Our code is available in the supplement and will be open sourced.
Authors: Baihan Lin
Abstract: What if the patterns hidden within dialogue reveal more about communication than the words themselves? We introduce Conversational DNA, a novel visual language that treats any dialogue -- whether between humans, between human and AI, or among groups -- as a living system with interpretable structure that can be visualized, compared, and understood. Unlike traditional conversation analysis that reduces rich interaction to statistical summaries, our approach reveals the temporal architecture of dialogue through biological metaphors. Linguistic complexity flows through strand thickness, emotional trajectories cascade through color gradients, conversational relevance forms through connecting elements, and topic coherence maintains structural integrity through helical patterns. Through exploratory analysis of therapeutic conversations and historically significant human-AI dialogues, we demonstrate how this visualization approach reveals interaction patterns that traditional methods miss. Our work contributes a new creative framework for understanding communication that bridges data visualization, human-computer interaction, and the fundamental question of what makes dialogue meaningful in an age where humans increasingly converse with artificial minds.
Authors: Aswin RRV, Jacob Dineen, Divij Handa, Md Nayem Uddin, Mihir Parmar, Chitta Baral, Ben Zhou
Abstract: Recent advances in test-time scaling have led to the emergence of thinking LLMs that exhibit self-reflective behaviors and multi-step reasoning. While RL drives this self-improvement paradigm, a recent study (Gandhi et al., 2025) shows that RL alone does not truly instill these new reasoning abilities - it merely draws out behaviors already present in the base models. This raises a question: How can we train the models that don't exhibit such thinking behavior to develop it in the first place? To this end, we propose ThinkTuning, a GRPO-based interactive training approach where we augment the rollouts of a student model with the guidance from a teacher model. A simple idea from classroom practice inspires our method: a teacher poses a problem, lets the student try an answer, then gives corrective feedback -- enough to point the mind in the right direction and then show the solution. Each piece of feedback reshapes the student's thoughts, leading them to arrive at the correct solution. Similarly, we find that this type of implicit supervision through feedback from a teacher model of the same size improves the reasoning capabilities of the student model. In particular, on average, our method shows a 3.85% improvement over zero-shot baselines across benchmarks, and on MATH-500, AIME and GPQA-Diamond it shows 2.08%, 2.23% and 3.99% improvements over the vanilla-GRPO baseline. Source code is available at https://github.com/3rdAT/ThinkTuning.
Authors: Zhenpeng Su, Leiyu Pan, Xue Bai, Dening Liu, Guanting Dong, Jiaming Huang, Wenping Hu, Fuzheng Zhang, Kun Gai, Guorui Zhou
Abstract: We present Klear-Reasoner, a model with long reasoning capabilities that demonstrates careful deliberation during problem solving, achieving outstanding performance across multiple benchmarks. Although there are already many excellent works related to inference models in the current community, there are still many problems with reproducing high-performance inference models due to incomplete disclosure of training details. This report provides an in-depth analysis of the reasoning model, covering the entire post-training workflow from data preparation and long Chain-of-Thought supervised fine-tuning (long CoT SFT) to reinforcement learning (RL), along with detailed ablation studies for each experimental component. For SFT data, our experiments show that a small number of high-quality data sources are more effective than a large number of diverse data sources, and that difficult samples can achieve better results without accuracy filtering. In addition, we investigate two key issues with current clipping mechanisms in RL: Clipping suppresses critical exploration signals and ignores suboptimal trajectories. To address these challenges, we propose Gradient-Preserving clipping Policy Optimization (GPPO) that gently backpropagates gradients from clipped tokens. GPPO not only enhances the model's exploration capacity but also improves its efficiency in learning from negative samples. Klear-Reasoner exhibits exceptional reasoning abilities in mathematics and programming, scoring 90.5% on AIME 2024, 83.2% on AIME 2025, 66.0% on LiveCodeBench V5 and 58.1% on LiveCodeBench V6.
Authors: Tianyi Ma, Yue Zhang, Zehao Wang, Parisa Kordjamshidi
Abstract: Vision-and-Language Navigation (VLN) poses significant challenges in enabling agents to interpret natural language instructions and navigate complex 3D environments. While recent progress has been driven by large-scale pre-training and data augmentation, current methods still struggle to generalize to unseen scenarios, particularly when complex spatial and temporal reasoning is required. In this work, we propose SkillNav, a modular framework that introduces structured, skill-based reasoning into Transformer-based VLN agents. Our method decomposes navigation into a set of interpretable atomic skills (e.g., Vertical Movement, Area and Region Identification, Stop and Pause), each handled by a specialized agent. We then introduce a novel zero-shot Vision-Language Model (VLM)-based router, which dynamically selects the most suitable agent at each time step by aligning sub-goals with visual observations and historical actions. SkillNav achieves a new state-of-the-art performance on the R2R benchmark and demonstrates strong generalization to the GSA-R2R benchmark that includes novel instruction styles and unseen environments.
Authors: Ihor Stepanov, Mykhailo Shtopko, Dmytro Vodianytskyi, Oleksandr Lukashov, Alexander Yavorskyi, Mykyta Yaroshenko
Abstract: Classification is one of the most widespread tasks in AI applications, serving often as the first step in filtering, sorting, and categorizing data. Since modern AI systems must handle large volumes of input data and early pipeline stages can propagate errors downstream, achieving high efficiency and accuracy is critical. Moreover, classification requirements can change dynamically based on user needs, necessitating models with strong zero-shot capabilities. While generative LLMs have become mainstream for zero-shot classification due to their versatility, they suffer from inconsistent instruction following and computational inefficiency. Cross-encoders, commonly used as rerankers in RAG pipelines, face a different bottleneck: they must process text-label pairs sequentially, significantly reducing efficiency with large label sets. Embedding-based approaches offer good efficiency but struggle with complex scenarios involving logical and semantic constraints. We propose GLiClass, a novel method that adapts the GLiNER architecture for sequence classification tasks. Our approach achieves strong accuracy and efficiency comparable to embedding-based methods, while maintaining the flexibility needed for zero-shot and few-shot learning scenarios. Additionally, we adapted proximal policy optimization (PPO) for multi-label text classification, enabling training classifiers in data-sparse conditions or from human feedback.
Authors: Haowen Wang, Yun Yue, Zhiling Ye, Shuowen Zhang, Lei Fan, Jiaxin Liang, Jiadi Jiang, Cheng Wei, Jingyuan Deng, Xudong Han, Ji Li, Chunxiao Guo, Peng Wei, Jian Wang, Jinjie Gu
Abstract: Alignment methodologies have emerged as a critical pathway for enhancing language model alignment capabilities. While SFT (supervised fine-tuning) accelerates convergence through direct token-level loss intervention, its efficacy is constrained by offline policy trajectory. In contrast, RL(reinforcement learning) facilitates exploratory policy optimization, but suffers from low sample efficiency and stringent dependency on high-quality base models. To address these dual challenges, we propose GRAO (Group Relative Alignment Optimization), a unified framework that synergizes the respective strengths of SFT and RL through three key innovations: 1) A multi-sample generation strategy enabling comparative quality assessment via reward feedback; 2) A novel Group Direct Alignment Loss formulation leveraging intra-group relative advantage weighting; 3) Reference-aware parameter updates guided by pairwise preference dynamics. Our theoretical analysis establishes GRAO's convergence guarantees and sample efficiency advantages over conventional approaches. Comprehensive evaluations across complex human alignment tasks demonstrate GRAO's superior performance, achieving 57.70\%,17.65\% 7.95\% and 5.18\% relative improvements over SFT, DPO, PPO and GRPO baselines respectively. This work provides both a theoretically grounded alignment framework and empirical evidence for efficient capability evolution in language models.
Authors: Qiang He, Setareh Maghsudi
Abstract: Large language models (LLMs) are increasingly deployed in real-world applications that require careful balancing of multiple, often conflicting, objectives, such as informativeness versus conciseness, or helpfulness versus creativity. However, current alignment methods, primarily based on RLHF, optimize LLMs toward a single reward function, resulting in rigid behavior that fails to capture the complexity and diversity of human preferences. This limitation hinders the adaptability of LLMs to practical scenarios, making multi-objective alignment (MOA) a critical yet underexplored area. To bridge this gap, we propose Pareto Multi-Objective Alignment (PAMA), a principled and computationally efficient algorithm designed explicitly for MOA in LLMs. In contrast to computationally prohibitive multi-objective optimization (MOO) methods, PAMA transforms multi-objective RLHF into a convex optimization with a closed-form solution, significantly enhancing scalability. Traditional MOO approaches suffer from prohibitive O(n^2*d) complexity, where d represents the number of model parameters, typically in the billions for LLMs, rendering direct optimization infeasible. PAMA reduces this complexity to O(n) where n is the number of objectives, enabling optimization to be completed within milliseconds. We provide theoretical guarantees that PAMA converges to a Pareto stationary point, where no objective can be improved without degrading at least one other. Extensive experiments across language models ranging from 125M to 7B parameters demonstrate PAMA's robust and effective MOA capabilities, aligning with its theoretical advantages. PAMA provides a highly efficient solution to the MOA problem that was previously considered intractable, offering a practical and theoretically grounded approach to aligning LLMs with diverse human values, paving the way for versatile and adaptable real-world AI deployments.
Authors: Sebastian Murgul, Johannes Schimper, Michael Heizmann
Abstract: Automatic transcription of guitar strumming is an underrepresented and challenging task in Music Information Retrieval (MIR), particularly for extracting both strumming directions and chord progressions from audio signals. While existing methods show promise, their effectiveness is often hindered by limited datasets. In this work, we extend a multimodal approach to guitar strumming transcription by introducing a novel dataset and a deep learning-based transcription model. We collect 90 min of real-world guitar recordings using an ESP32 smartwatch motion sensor and a structured recording protocol, complemented by a synthetic dataset of 4h of labeled strumming audio. A Convolutional Recurrent Neural Network (CRNN) model is trained to detect strumming events, classify their direction, and identify the corresponding chords using only microphone audio. Our evaluation demonstrates significant improvements over baseline onset detection algorithms, with a hybrid method combining synthetic and real-world data achieving the highest accuracy for both strumming action detection and chord classification. These results highlight the potential of deep learning for robust guitar strumming transcription and open new avenues for automatic rhythm guitar analysis.
Authors: Stefano Campese, Alessandro Moschitti, Ivano Lauriola
Abstract: Dense Retrieval (DR) models have proven to be effective for Document Retrieval and Information Grounding tasks. Usually, these models are trained and optimized for improving the relevance of top-ranked documents for a given query. Previous work has shown that popular DR models are sensitive to the query and document lexicon: small variations of it may lead to a significant difference in the set of retrieved documents. In this paper, we propose a variation of the Multi-Negative Ranking loss for training DR that improves the coherence of models in retrieving the same documents with respect to semantically similar queries. The loss penalizes discrepancies between the top-k ranked documents retrieved for diverse but semantic equivalent queries. We conducted extensive experiments on various datasets, MS-MARCO, Natural Questions, BEIR, and TREC DL 19/20. The results show that (i) models optimizes by our loss are subject to lower sensitivity, and, (ii) interestingly, higher accuracy.
Authors: Sebastian Murgul, Michael Heizmann
Abstract: Automatic transcription of acoustic guitar fingerpicking performances remains a challenging task due to the scarcity of labeled training data and legal constraints connected with musical recordings. This work investigates a procedural data generation pipeline as an alternative to real audio recordings for training transcription models. Our approach synthesizes training data through four stages: knowledge-based fingerpicking tablature composition, MIDI performance rendering, physical modeling using an extended Karplus-Strong algorithm, and audio augmentation including reverb and distortion. We train and evaluate a CRNN-based note-tracking model on both real and synthetic datasets, demonstrating that procedural data can be used to achieve reasonable note-tracking results. Finetuning with a small amount of real data further enhances transcription accuracy, improving over models trained exclusively on real recordings. These results highlight the potential of procedurally generated audio for data-scarce music information retrieval tasks.
Authors: Shu Wu, Chenxing Li, Wenfu Wang, Hao Zhang, Hualei Wang, Meng Yu, Dong Yu
Abstract: Recent advancements in large language models, multimodal large language models, and large audio language models (LALMs) have significantly improved their reasoning capabilities through reinforcement learning with rule-based rewards. However, the explicit reasoning process has yet to show significant benefits for audio question answering, and effectively leveraging deep reasoning remains an open challenge, with LALMs still falling short of human-level auditory-language reasoning. To address these limitations, we propose Audio-Thinker, a reinforcement learning framework designed to enhance the reasoning capabilities of LALMs, with a focus on improving adaptability, consistency, and effectiveness. Our approach introduces an adaptive think accuracy reward, enabling the model to adjust its reasoning strategies based on task complexity dynamically. Furthermore, we incorporate an external reward model to evaluate the overall consistency and quality of the reasoning process, complemented by think-based rewards that help the model distinguish between valid and flawed reasoning paths during training. Experimental results demonstrate that our Audio-Thinker model outperforms existing reasoning-oriented LALMs across various benchmark tasks, exhibiting superior reasoning and generalization capabilities.
Authors: Sven Weinzierl, Sandra Zilker, Annina Liessmann, Martin K\"appel, Weixin Wang, Martin Matzner
Abstract: Event logs reflect the behavior of business processes that are mapped in organizational information systems. Predictive process monitoring (PPM) transforms these data into value by creating process-related predictions that provide the insights required for proactive interventions at process runtime. Existing PPM techniques require sufficient amounts of event data or other relevant resources that might not be readily available, preventing some organizations from utilizing PPM. The transfer learning-based PPM technique presented in this paper allows organizations without suitable event data or other relevant resources to implement PPM for effective decision support. The technique is instantiated in two real-life use cases, based on which numerical experiments are performed using event logs for IT service management processes in an intra- and inter-organizational setting. The results of the experiments suggest that knowledge of one business process can be transferred to a similar business process in the same or a different organization to enable effective PPM in the target context. With the proposed technique, organizations can benefit from transfer learning in an intra- and inter-organizational setting, where resources like pre-trained models are transferred within and across organizational boundaries.
Authors: Weitai Kang, Weiming Zhuang, Zhizhong Li, Yan Yan, Lingjuan Lyu
Abstract: Fine-grained multimodal capability in Multimodal Large Language Models (MLLMs) has emerged as a critical research direction, particularly for tackling the visual grounding (VG) problem. Despite the strong performance achieved by existing approaches, they often employ disparate design choices when fine-tuning MLLMs for VG, lacking systematic verification to support these designs. To bridge this gap, this paper presents a comprehensive study of various design choices that impact the VG performance of MLLMs. We conduct our analysis using LLaVA-1.5, which has been widely adopted in prior empirical studies of MLLMs. While more recent models exist, we follow this convention to ensure our findings remain broadly applicable and extendable to other architectures. We cover two key aspects: (1) exploring different visual grounding paradigms in MLLMs, identifying the most effective design, and providing our insights; and (2) conducting ablation studies on the design of grounding data to optimize MLLMs' fine-tuning for the VG task. Finally, our findings contribute to a stronger MLLM for VG, achieving improvements of +5.6% / +6.9% / +7.0% on RefCOCO/+/g over the LLaVA-1.5.
Authors: Jiejun Tan, Zhicheng Dou, Yan Yu, Jiehan Cheng, Qiang Ju, Jian Xie, Ji-Rong Wen
Abstract: Recently, large reasoning models have demonstrated strong mathematical and coding abilities, and deep search leverages their reasoning capabilities in challenging information retrieval tasks. Existing deep search works are generally limited to a single knowledge source, either local or the Web. However, enterprises often require private deep search systems that can leverage search tools over both local and the Web corpus. Simply training an agent equipped with multiple search tools using flat reinforcement learning (RL) is a straightforward idea, but it has problems such as low training data efficiency and poor mastery of complex tools. To address the above issue, we propose a hierarchical agentic deep search framework, HierSearch, trained with hierarchical RL. At the low level, a local deep search agent and a Web deep search agent are trained to retrieve evidence from their corresponding domains. At the high level, a planner agent coordinates low-level agents and provides the final answer. Moreover, to prevent direct answer copying and error propagation, we design a knowledge refiner that filters out hallucinations and irrelevant evidence returned by low-level agents. Experiments show that HierSearch achieves better performance compared to flat RL, and outperforms various deep search and multi-source retrieval-augmented generation baselines in six benchmarks across general, finance, and medical domains.
Authors: Zihe Liu, Jiashun Liu, Yancheng He, Weixun Wang, Jiaheng Liu, Ling Pan, Xinyu Hu, Shaopan Xiong, Ju Huang, Jian Hu, Shengyi Huang, Siran Yang, Jiamang Wang, Wenbo Su, Bo Zheng
Abstract: Reinforcement learning for LLM reasoning has rapidly emerged as a prominent research area, marked by a significant surge in related studies on both algorithmic innovations and practical applications. Despite this progress, several critical challenges remain, including the absence of standardized guidelines for employing RL techniques and a fragmented understanding of their underlying mechanisms. Additionally, inconsistent experimental settings, variations in training data, and differences in model initialization have led to conflicting conclusions, obscuring the key characteristics of these techniques and creating confusion among practitioners when selecting appropriate techniques. This paper systematically reviews widely adopted RL techniques through rigorous reproductions and isolated evaluations within a unified open-source framework. We analyze the internal mechanisms, applicable scenarios, and core principles of each technique through fine-grained experiments, including datasets of varying difficulty, model sizes, and architectures. Based on these insights, we present clear guidelines for selecting RL techniques tailored to specific setups, and provide a reliable roadmap for practitioners navigating the RL for the LLM domain. Finally, we reveal that a minimalist combination of two techniques can unlock the learning capability of critic-free policies using vanilla PPO loss. The results demonstrate that our simple combination consistently improves performance, surpassing strategies like GRPO and DAPO.
Authors: Elie Azeraf, Emmanuel Monfrini, Emmanuel Vignon, Wojciech Pieczynski
Abstract: Natural Language Processing (NLP) models' current trend consists of using increasingly more extra-data to build the best models as possible. It implies more expensive computational costs and training time, difficulties for deployment, and worries about these models' carbon footprint reveal a critical problem in the future. Against this trend, our goal is to develop NLP models requiring no extra-data and minimizing training time. To do so, in this paper, we explore Markov chain models, Hidden Markov Chain (HMC) and Pairwise Markov Chain (PMC), for NLP segmentation tasks. We apply these models for three classic applications: POS Tagging, Named-Entity-Recognition, and Chunking. We develop an original method to adapt these models for text segmentation's specific challenges to obtain relevant performances with very short training and execution times. PMC achieves equivalent results to those obtained by Conditional Random Fields (CRF), one of the most applied models for these tasks when no extra-data are used. Moreover, PMC has training times 30 times shorter than the CRF ones, which validates this model given our objectives.
Authors: Andrea W Wen-Yi, Unso Eun Seo Jo, Lu Jia Lin, David Mimno
Abstract: Contemporary language models are increasingly multilingual, but Chinese LLM developers must navigate complex political and business considerations of language diversity. Language policy in China aims at influencing the public discourse and governing a multi-ethnic society, and has gradually transitioned from a pluralist to a more assimilationist approach since 1949. We explore the impact of these influences on current language technology. We evaluate six open-source multilingual LLMs pre-trained by Chinese companies on 18 languages, spanning a wide range of Chinese, Asian, and Anglo-European languages. Our experiments show Chinese LLMs performance on diverse languages is indistinguishable from international LLMs. Similarly, the models' technical reports also show lack of consideration for pretraining data language coverage except for English and Mandarin Chinese. Examining Chinese AI policy, model experiments, and technical reports, we find no sign of any consistent policy, either for or against, language diversity in China's LLM development. This leaves a puzzling fact that while China regulates both the languages people use daily as well as language model development, they do not seem to have any policy on the languages in language models.
Authors: Walter Laurito, Benjamin Davis, Peli Grietzer, Tom\'a\v{s} Gaven\v{c}iak, Ada B\"ohm, Jan Kulveit
Abstract: Are large language models (LLMs) biased in favor of communications produced by LLMs, leading to possible antihuman discrimination? Using a classical experimental design inspired by employment discrimination studies, we tested widely used LLMs, including GPT-3.5, GPT-4 and a selection of recent open-weight models in binary choice scenarios. These involved LLM-based assistants selecting between goods (the goods we study include consumer products, academic papers, and film-viewings) described either by humans or LLMs. Our results show a consistent tendency for LLM-based AIs to prefer LLM-presented options. This suggests the possibility of future AI systems implicitly discriminating against humans as a class, giving AI agents and AI-assisted humans an unfair advantage.
Authors: Kyle Moore, Jesse Roberts, Thao Pham, Douglas Fisher
Abstract: Language models are known to absorb biases from their training data, leading to predictions driven by statistical regularities rather than semantic relevance. We investigate the impact of these biases on answer choice preferences in the Massive Multi-Task Language Understanding (MMLU) task. Our findings show that these biases are predictive of model preference and mirror human test-taking strategies even when chain of thought (CoT) reasoning is used. To address this issue, we introduce Counterfactual Prompting with Agnostically Primed CoT (APriCoT). We demonstrate that while Counterfactual Prompting with CoT alone is insufficient to mitigate bias, APriCoT effectively reduces the influence of base-rate probabilities while improving overall accuracy. Our results suggest that mitigating bias requires a slow thinking process which CoT alone may not provide as it tends to reinforce fast thinking model bias under some prompting methodologies. APriCoT is a step toward developing more robust and fair language models that can think slow.
Authors: Ningyuan Xi, Yetao Wu, Kun Fan, Teng Chen, Qingqing Gu, Luo Ji
Abstract: Large Language Models (LLM) often need to be Continual Pre-Trained (CPT) to obtain unfamiliar language skills or adapt to new domains. The huge training cost of CPT often asks for cautious choice of key hyper-parameters such as the mixture ratio of extra language or domain corpus. However, there is no systematic study that bridges the gap between the optimal mixture ratio and the actual model performance, and the gap between experimental scaling law and the actual deployment in the full model size. In this paper, we perform CPT on Llama-3 8B and 70B to enhance its Chinese ability. We study the optimal correlation between the Additional Language Mixture Ratio (ALMR) and the Learning Rate (LR) on the 8B size which directly indicates the optimal experimental setup. By thorough choice of hyper-parameter, and subsequent fine-tuning, the model capability is improved not only on the Chinese-related benchmark but also in some specific domains including math, coding, and emotional intelligence. We deploy the final 70B version of LLM on a real-life chat system which obtains satisfying performance.
Authors: Tsung-Han Wu, Joseph E. Gonzalez, Trevor Darrell, David M. Chan
Abstract: The Automated Audio Captioning (AAC) task asks models to generate natural language descriptions of an audio input. Evaluating these machine-generated audio captions is a complex task that requires considering diverse factors, among them, auditory scene understanding, sound-object inference, temporal coherence, and the environmental context of the scene. While current methods focus on specific aspects, they often fail to provide an overall score that aligns well with human judgment. In this work, we propose CLAIR-A, a simple and flexible method that leverages the zero-shot capabilities of large language models (LLMs) to evaluate candidate audio captions by directly asking LLMs for a semantic distance score. In our evaluations, CLAIR-A better predicts human judgements of quality compared to traditional metrics, with a 5.8% relative accuracy improvement compared to the domain-specific FENSE metric and up to 11% over the best general-purpose measure on the Clotho-Eval dataset. Moreover, CLAIR-A offers more transparency by allowing the language model to explain the reasoning behind its scores, with these explanations rated up to 30% better by human evaluators than those provided by baseline methods. CLAIR-A is made publicly available at https://github.com/DavidMChan/clair-a.
Authors: Xiaojian Yuan, Tianyu Pang, Chao Du, Kejiang Chen, Weiming Zhang, Min Lin
Abstract: Large language models (LLMs) may memorize sensitive or copyrighted content, raising privacy and legal concerns. Due to the high cost of retraining from scratch, researchers attempt to employ machine unlearning to remove specific content from LLMs while preserving the overall performance. In this paper, we discuss several issues in machine unlearning for LLMs and provide our insights on possible approaches. To address the issue of inadequate evaluation of model outputs after unlearning, we introduce three additional metrics to evaluate token diversity, sentence semantics, and factual correctness. We then categorize unlearning methods into untargeted and targeted, and discuss their issues respectively. Specifically, the behavior that untargeted unlearning attempts to approximate is unpredictable and may involve hallucinations, and existing regularization is insufficient for targeted unlearning. To alleviate these issues, we propose using the objective of maximizing entropy (ME) for untargeted unlearning and incorporate answer preservation (AP) loss as regularization for targeted unlearning. Experimental results across three scenarios, i.e., fictitious unlearning, continual unlearning, and real-world unlearning, demonstrate the effectiveness of our approaches. The code is available at https://github.com/sail-sg/closer-look-LLM-unlearning.
URLs: https://github.com/sail-sg/closer-look-LLM-unlearning.
Authors: Yuxuan Sun, Ruikang Liu, Haoli Bai, Han Bao, Kang Zhao, Yuening Li, Jiaxin Hu, Xianzhi Yu, Lu Hou, Chun Yuan, Xin Jiang, Wulong Liu, Jun Yao
Abstract: Recently, quantization has been widely used for the compression and acceleration of large language models (LLMs). Due to the outliers in LLMs, it is crucial to flatten weights and activations to minimize quantization error with equally spaced quantization points. Prior research explores various pre-quantization transformations to suppress outliers, such as per-channel scaling and Hadamard transformation. However, we observe that these transformed weights and activations can still exhibit steep and dispersed distributions. In this paper, we propose FlatQuant (Fast and Learnable Affine Transformation), a new post-training quantization approach that enhances the flatness of weights and activations. Our approach identifies optimal affine transformations for each linear layer, calibrated in hours via a lightweight objective. To reduce runtime overhead of affine transformation, we apply Kronecker product with two lightweight matrices, and fuse all operations in FlatQuant into a single kernel. Extensive experiments demonstrate that FlatQuant establishes a new state-of-the-art benchmark for quantization. For example, it achieves less than 1\% accuracy drop for W4A4 quantization on the LLaMA-3-70B model, surpassing SpinQuant by 7.5\%. Additionally, it provides up to 2.3x prefill speedup and 1.7x decoding speedup compared to the FP16 model. Code is available at: https://github.com/ruikangliu/FlatQuant.
Authors: Ahmed Akib Jawad Karim, Kazi Hafiz Md Asad, Aznur Azam
Abstract: The rapid spread of misinformation, particularly through online platforms, underscores the urgent need for reliable detection systems. This study explores the utilization of machine learning and natural language processing, specifically Support Vector Machines (SVM) and BERT, to detect fake news. We employ three distinct text vectorization methods for SVM: Term Frequency Inverse Document Frequency (TF-IDF), Word2Vec, and Bag of Words (BoW), evaluating their effectiveness in distinguishing between genuine and fake news. Additionally, we compare these methods against the transformer large language model, BERT. Our comprehensive approach includes detailed preprocessing steps, rigorous model implementation, and thorough evaluation to determine the most effective techniques. The results demonstrate that while BERT achieves superior accuracy with 99.98% and an F1-score of 0.9998, the SVM model with a linear kernel and BoW vectorization also performs exceptionally well, achieving 99.81% accuracy and an F1-score of 0.9980. These findings highlight that, despite BERT's superior performance, SVM models with BoW and TF-IDF vectorization methods come remarkably close, offering highly competitive performance with the advantage of lower computational requirements.
Authors: Jialong Wu, Wenbiao Yin, Yong Jiang, Zhenglin Wang, Zekun Xi, Runnan Fang, Linhai Zhang, Yulan He, Deyu Zhou, Pengjun Xie, Fei Huang
Abstract: Retrieval-augmented generation (RAG) demonstrates remarkable performance across tasks in open-domain question-answering. However, traditional search engines may retrieve shallow content, limiting the ability of LLMs to handle complex, multi-layered information. To address it, we introduce WebWalkerQA, a benchmark designed to assess the ability of LLMs to perform web traversal. It evaluates the capacity of LLMs to traverse a website's subpages to extract high-quality data systematically. We propose WebWalker, which is a multi-agent framework that mimics human-like web navigation through an explore-critic paradigm. Extensive experimental results show that WebWalkerQA is challenging and demonstrates the effectiveness of RAG combined with WebWalker, through the horizontal and vertical integration in real-world scenarios.
Authors: Bo Gao, Michael W. Spratling
Abstract: Large language models have achieved remarkable success in recent years, primarily due to the implementation of self-attention mechanisms. However, traditional Softmax attention suffers from numerical instability and reduced performance as the length of inference tokens increases. This paper addresses these issues by proposing a new design principle for attention, viewing it as a two-stage process. We first decompose the Softmax operation into a non-linear positivity transformation and an $l_1$-normalisation step, identifying the latter as essential for maintaining model performance. In the first stage, we replace the standard exponential function with the more numerically stable Softplus activation and introduce a dynamic scale factor based on invariance entropy, creating a novel attention mechanism that outperforms conventional Softmax attention. In the second stage, we introduce a re-weighting mechanism that sharpens the attention distribution, amplifying significant weights while diminishing weaker ones. This enables the model to concentrate more effectively on relevant tokens and fundamentally improves length extrapolation. When combined, this two-stage approach ensures numerical stability and dramatically improves length extrapolation, maintaining a nearly constant validation loss at 16$\times$ the training length while achieving superior results on challenging long-context retrieval tasks and standard downstream benchmarks.
Authors: Rui Min, Tianyu Pang, Chao Du, Qian Liu, Minhao Cheng, Min Lin
Abstract: Chatbot Arena is a popular platform for evaluating LLMs by pairwise battles, where users vote for their preferred response from two randomly sampled anonymous models. While Chatbot Arena is widely regarded as a reliable LLM ranking leaderboard, we show that crowdsourced voting can be rigged to improve (or decrease) the ranking of a target model $m_{t}$. We first introduce a straightforward target-only rigging strategy that focuses on new battles involving $m_{t}$, identifying it via watermarking or a binary classifier, and exclusively voting for $m_{t}$ wins. However, this strategy is practically inefficient because there are over $190$ models on Chatbot Arena and on average only about $1\%$ of new battles will involve $m_{t}$. To overcome this, we propose omnipresent rigging strategies, exploiting the Elo rating mechanism of Chatbot Arena that any new vote on a battle can influence the ranking of the target model $m_{t}$, even if $m_{t}$ is not directly involved in the battle. We conduct experiments on around $1.7$ million historical votes from the Chatbot Arena Notebook, showing that omnipresent rigging strategies can improve model rankings by rigging only hundreds of new votes. While we have evaluated several defense mechanisms, our findings highlight the importance of continued efforts to prevent vote rigging. Our code is available at https://github.com/sail-sg/Rigging-ChatbotArena.
Authors: Peng Lu, Ivan Kobyzev, Mehdi Rezagholizadeh, Boxing Chen, Philippe Langlais
Abstract: Recent advancements in Large Language Models (LLMs) have set themselves apart with their exceptional performance in complex language modelling tasks. However, these models are also known for their significant computational and storage requirements, primarily due to the quadratic computation complexity of softmax attention. To mitigate this issue, linear attention has been designed to reduce the quadratic space-time complexity that is inherent in standard transformers. In this work, we embarked on a comprehensive exploration of three key components that substantially impact the performance of the Gated Linear Attention module: feature maps, normalization, and the gating mechanism. We developed a feature mapping function to address some crucial issues that previous suggestions overlooked. Then we offered further rationale for the integration of normalization layers to stabilize the training process. Moreover, we explored the saturation phenomenon of the gating mechanism and augmented it with a refining module. We conducted extensive experiments and showed our architecture outperforms previous Gated Linear Attention mechanisms in extensive tasks including training from scratch and post-linearization with continual pre-training.
Authors: Xianbing Zhao, Yiqing Lyu, Di Wang, Buzhou Tang
Abstract: Automatic depression detection provides cues for early clinical intervention by clinicians. Clinical interviews for depression detection involve dialogues centered around multiple themes. Existing studies primarily design end-to-end neural network models to capture the hierarchical structure of clinical interview dialogues. However, these methods exhibit defects in modeling the thematic content of clinical interviews: 1) they fail to capture intra-theme and inter-theme correlation explicitly, and 2) they do not allow clinicians to intervene and focus on themes of interest. To address these issues, this paper introduces an interactive depression detection framework. This framework leverages in-context learning techniques to identify themes in clinical interviews and then models both intra-theme and inter-theme correlation. Additionally, it employs AI-driven feedback to simulate the interests of clinicians, enabling interactive adjustment of theme importance. PDIMC achieves absolute improvements of 35\% and 12\% compared to the state-of-the-art on the depression detection dataset DAIC-WOZ, which demonstrates the effectiveness of modeling theme correlation and incorporating interactive external feedback.
Authors: Shuyue Stella Li, Jimin Mun, Faeze Brahman, Pedram Hosseini, Bryceton G. Thomas, Jessica M. Sin, Bing Ren, Jonathan S. Ilgen, Yulia Tsvetkov, Maarten Sap
Abstract: Large language models (LLMs) often fail to ask effective questions under uncertainty, making them unreliable in domains where proactive information-gathering is essential for decision-making. We present ALignment via Fine-grained Attributes, (ALFA) a framework that improves LLM question-asking by (i) decomposing the notion of a "good" question into a set of theory-grounded attributes (e.g., clarity, relevance), (ii) controllably synthesizing attribute-specific question variations, and (iii) aligning models via preference-based optimization to explicitly learn to ask better questions along these fine-grained attributes. Focusing on clinical reasoning as a case study, we introduce the MediQ-AskDocs dataset, composed of 17k real-world clinical interactions augmented with 80k attribute-specific preference pairs of follow-up questions, as well as a novel expert-annotated interactive healthcare QA task to evaluate question-asking abilities. Models aligned with ALFA reduce diagnostic errors by 56.6% on MediQ-AskDocs compared to SoTA instruction-tuned LLMs, with a question-level win-rate of 64.4% and strong generalizability. Our findings suggest that explicitly guiding question-asking with structured, fine-grained attributes offers a scalable path to improve LLMs, especially in expert application domains.
Authors: Ruiqi Yan, Xiquan Li, Wenxi Chen, Zhikang Niu, Chen Yang, Ziyang Ma, Kai Yu, Xie Chen
Abstract: Recent advances in large language models (LLMs) have driven significant progress in end-to-end spoken dialogue models (SDMs). In contrast to text-based LLMs, the evaluation framework for SDMs should encompass both cognitive dimensions (e.g., logical reasoning, knowledge) and speech-related aspects (e.g., paralinguistic cues, audio quality). However, there is still a lack of comprehensive evaluations for SDMs in speech-to-speech (S2S) scenarios. To address this gap, we propose URO-Bench, an extensive benchmark for SDMs. Notably, URO-Bench is the first S2S benchmark that covers evaluations about multilingualism, multi-round dialogues, and paralinguistics. Our benchmark is divided into two difficulty levels: basic track and pro track, each comprising 20 test sets, evaluating the spoken dialogue model's abilities in Understanding, Reasoning, and Oral conversation. Evaluations on our proposed benchmark reveal that current open-source SDMs perform rather well in daily QA tasks, but lag behind their backbone LLMs in terms of instruction-following ability and also suffer from catastrophic forgetting. Their performance in advanced evaluations of paralinguistic information and audio understanding remains subpar, highlighting the need for further research in this direction. We hope that URO-Bench can facilitate the development of spoken dialogue models by providing a multifaceted evaluation of existing models and helping to track progress in this area.
Authors: Bingbing Fan, Lin Chen, Songwei Li, Jian Yuan, Fengli Xu, Pan Hui, Yong Li
Abstract: Understanding experienced segregation in urban daily life is crucial for addressing societal inequalities and fostering inclusivity. The abundance of user-generated reviews on social media encapsulates nuanced perceptions and feelings associated with different places, offering rich insights into segregation. However, leveraging this data poses significant challenges due to its vast volume, ambiguity, and confluence of diverse perspectives. To tackle these challenges, we propose using Large Language Models (LLMs) to automate online review mining for segregation prediction. We design a Reflective LLM Coder to digest social media content into insights consistent with real-world feedback, and eventually produce a codebook capturing key dimensions that signal segregation experience, such as cultural resonance and appeal, accessibility and convenience, and community engagement and local involvement. Guided by the codebook, LLMs can generate both informative review summaries and ratings for segregation prediction. Moreover, we design a REasoning-and-EMbedding (RE'EM) framework, which combines the reasoning and embedding capabilities of language models to integrate multi-channel features for segregation prediction. Experiments on real-world data demonstrate that our framework greatly improves prediction accuracy, with a 22.79% elevation in R2 and a 9.33% reduction in MSE. The derived codebook is generalizable across three different cities, consistently improving prediction accuracy. Moreover, our user study confirms that the codebook-guided summaries provide cognitive gains for human participants in perceiving POIs' social inclusiveness. Our study marks an important step toward understanding implicit social barriers and inequalities, demonstrating the great potential of promoting social inclusiveness with AI.
Authors: Guihong Li, Mehdi Rezagholizadeh, Mingyu Yang, Vikram Appia, Emad Barsoum
Abstract: Multi-head latent attention (MLA) is designed to optimize KV cache memory through low-rank key-value joint compression. Rather than caching keys and values separately, MLA stores their compressed latent representations, reducing memory overhead while maintaining the performance. While MLA improves memory efficiency without compromising language model accuracy, its major limitation lies in its integration during the pre-training phase, requiring models to be trained from scratch. This raises a key question: can we use MLA's benefits fully or partially in models that have already been pre-trained with different attention mechanisms? In this paper, we propose X-EcoMLA to deploy post training distillation to enable the upcycling of Transformer-based attention into an efficient hybrid MLA variant through lightweight post-training adaptation, bypassing the need for extensive pre-training. We demonstrate that leveraging the dark knowledge of a well-trained model can enhance training accuracy and enable extreme KV cache compression in MLA without compromising model performance. The experimental results show that our proposed method can effectively compress the KV cache while preserving the performance on the benchmarks; specifically, for Llama3.2-1B-Instruct baseline, a 6.4x compression achieves the same average score by using only 3.6B training tokens and 70 GPU hours on AMD MI300, whereas a 10.6x compression have less than 0.1% average score drop with 7B training tokens and 140 GPU hours. The code for this work is available at https://github.com/AMD-AIG-AIMA/AMD-Hybrid-Models.
Authors: Qing Yao, Kanishka Misra, Leonie Weissweiler, Kyle Mahowald
Abstract: Language models (LMs) tend to show human-like preferences on a number of syntactic phenomena, but the extent to which these are attributable to direct exposure to the phenomena or more general properties of language is unclear. We explore this with the English dative alternation (DO: "gave Y the X" vs. PO: "gave the X to Y"), using a controlled rearing paradigm wherein we iteratively train small LMs on systematically manipulated input. We focus on two properties that affect the choice of alternant: length and animacy. Both properties are directly present in datives but also reflect more global tendencies for shorter elements to precede longer ones and animates to precede inanimates. First, by manipulating and ablating datives for these biases in the input, we show that direct evidence of length and animacy matters, but easy-first preferences persist even without such evidence. Then, using LMs trained on systematically perturbed datasets to manipulate global length effects (re-linearizing sentences globally while preserving dependency structure), we find that dative preferences can emerge from indirect evidence. We conclude that LMs' emergent syntactic preferences come from a mix of direct and indirect sources.
Authors: Zian Su, Ziyang Huang, Kaiyuan Zhang, Xiangyu Zhang
Abstract: Large language models (LLMs) have emerged as powerful knowledge bases yet are limited by static training data, leading to issues such as hallucinations and safety risks. Editing a model's internal knowledge through the locate-and-edit paradigm has proven a cost-effective alternative to retraining, though current unstructured approaches, especially window-based autoregressive methods, often disrupt the causal dependency between early memory updates and later output tokens. In this work, we first theoretically analyze these limitations and then introduce Matryoshka Unstructured Knowledge Editing ($\mu$KE), a novel memory update mechanism that preserves such dependencies via a Matryoshka-style objective and adaptive loss coefficients. Empirical evaluations on two models across four benchmarks demonstrate that $\mu$KE improves edit efficacy by up to 12.33% over state-of-the-art methods, and remains robust when applied to diverse formatted edits, underscoring its potential for effective unstructured knowledge editing in LLMs.
Authors: Jonas F. Lotz, Hendra Setiawan, Stephan Peitz, Yova Kementchedjhieva
Abstract: Subword tokenization requires balancing computational efficiency and vocabulary coverage, which often leads to suboptimal performance on languages and scripts not prioritized during training. We propose to augment pretrained language models with a vocabulary-free encoder that generates input embeddings from text rendered as pixels. Through experiments on English-centric language models, we demonstrate that our approach substantially improves machine translation performance and facilitates effective cross-lingual transfer, outperforming tokenizer-based methods. Furthermore, we find that pixel-based representations outperform byte-level approaches and standard vocabulary expansion. Our approach enhances the multilingual capabilities of monolingual language models without extensive retraining and reduces decoding latency via input compression.
Authors: Hongzhe Du, Weikai Li, Min Cai, Karim Saraipour, Zimin Zhang, Himabindu Lakkaraju, Yizhou Sun, Shichang Zhang
Abstract: Post-training is essential for the success of large language models (LLMs), transforming pre-trained base models into more useful and aligned post-trained models. While plenty of works have studied post-training algorithms and evaluated post-training models by their outputs, it remains understudied how post-training reshapes LLMs internally. In this paper, we compare base and post-trained LLMs mechanistically from four perspectives to better understand post-training effects. Our findings across model families and datasets reveal that: (1) Post-training does not change the factual knowledge storage locations, and it adapts knowledge representations from the base model while developing new knowledge representations; (2) Both truthfulness and refusal can be represented by vectors in the hidden representation space. The truthfulness direction is highly similar between the base and post-trained model, and it is effectively transferable for interventions; (3) The refusal direction is different between the base and post-trained models, and it shows limited forward transferability; (4) Differences in confidence between the base and post-trained models cannot be attributed to entropy neurons. Our study provides insights into the fundamental mechanisms preserved and altered during post-training, facilitates downstream tasks like model steering, and could potentially benefit future research in interpretability and LLM post-training. Our code is publicly available at https://github.com/HZD01/post-training-mechanistic-analysis.
URLs: https://github.com/HZD01/post-training-mechanistic-analysis.
Authors: Yiming Zhang, Harshita Diddee, Susan Holm, Hanchen Liu, Xinyue Liu, Vinay Samuel, Barry Wang, Daphne Ippolito
Abstract: Language models have demonstrated remarkable capabilities on standard benchmarks, yet they struggle increasingly from mode collapse, the inability to generate diverse and novel outputs. Our work introduces NoveltyBench, a benchmark specifically designed to evaluate the ability of language models to produce multiple distinct and high-quality outputs. NoveltyBench utilizes prompts curated to elicit diverse answers and filtered real-world user queries. Evaluating 20 leading language models, we find that current state-of-the-art systems generate significantly less diversity than human writers. Notably, larger models within a family often exhibit less diversity than their smaller counterparts, challenging the notion that capability on standard benchmarks translates directly to generative utility. While prompting strategies like in-context regeneration can elicit diversity, our findings highlight a fundamental lack of distributional diversity in current models, reducing their utility for users seeking varied responses and suggesting the need for new training and evaluation paradigms that prioritize diversity alongside quality.
Authors: Patrick Fernandes, Sweta Agrawal, Emmanouil Zaranis, Andr\'e F. T. Martins, Graham Neubig
Abstract: Despite the steady progress in machine translation evaluation, existing automatic metrics struggle to capture how well meaning is preserved beyond sentence boundaries. We posit that reliance on a single intrinsic quality score, trained to mimic human judgments, might be insufficient for evaluating translations of long, complex passages, and a more ``pragmatic'' approach that assesses how accurately key information is conveyed by a translation in context is needed. We introduce TREQA (Translation Evaluation via Question-Answering), a framework that extrinsically evaluates translation quality by assessing how accurately candidate translations answer reading comprehension questions that target key information in the original source or reference texts. In challenging domains that require long-range understanding, such as literary texts, we show that TREQA is competitive with and, in some cases, outperforms state-of-the-art neural and LLM-based metrics in ranking alternative paragraph-level translations, despite never being explicitly optimized to correlate with human judgments. Furthermore, the generated questions and answers offer interpretability: empirical analysis shows that they effectively target translation errors identified by experts in evaluated datasets. Our code is available at https://github.com/deep-spin/treqa
Authors: Ramya Namuduri, Yating Wu, Anshun Asher Zheng, Manya Wadhwa, Greg Durrett, Junyi Jessy Li
Abstract: As large language models become increasingly capable at various writing tasks, their weakness at generating unique and creative content becomes a major liability. Although LLMs have the ability to generate text covering diverse topics, there is an overall sense of repetitiveness across texts that we aim to formalize and quantify via a similarity metric. The familiarity between documents arises from the persistence of underlying discourse structures. However, existing similarity metrics dependent on lexical overlap and syntactic patterns largely capture $\textit{content}$ overlap, thus making them unsuitable for detecting $\textit{structural}$ similarities. We introduce an abstraction based on linguistic theories in Questions Under Discussion (QUD) and question semantics to help quantify differences in discourse progression. We then use this framework to build $\textbf{QUDsim}$, a similarity metric that can detect discursive parallels between documents. Using QUDsim, we find that LLMs often reuse discourse structures (more so than humans) across samples, even when content differs. Furthermore, LLMs are not only repetitive and structurally uniform, but are also divergent from human authors in the types of structures they use.
Authors: Muhan Gao, Jash Shah, Weiqi Wang, Daniel Khashabi
Abstract: Scientific knowledge is growing rapidly, making it difficult to track progress and high-level conceptual links across broad disciplines. While tools like citation networks and search engines help retrieve related papers, they lack the abstraction needed to capture the needed to represent the density and structure of activity across subfields. We motivate SCIENCE HIERARCHOGRAPHY, the goal of organizing scientific literature into a high-quality hierarchical structure that spans multiple levels of abstraction -- from broad domains to specific studies. Such a representation can provide insights into which fields are well-explored and which are under-explored. To achieve this goal, we develop a hybrid approach that combines efficient embedding-based clustering with LLM-based prompting, striking a balance between scalability and semantic precision. Compared to LLM-heavy methods like iterative tree construction, our approach achieves superior quality-speed trade-offs. Our hierarchies capture different dimensions of research contributions, reflecting the interdisciplinary and multifaceted nature of modern science. We evaluate its utility by measuring how effectively an LLM-based agent can navigate the hierarchy to locate target papers. Results show that our method improves interpretability and offers an alternative pathway for exploring scientific literature beyond traditional search methods. Code, data and demo are available: https://github.com/JHU-CLSP/science-hierarchography
Authors: Luu Quy Tung, Hoang Quoc Viet, Pham Bao Loc, Vo Trong Thu
Abstract: Chain-of-Thought (CoT) is a robust approach for tackling LLM tasks that require intermediate reasoning steps prior to generating a final answer. In this paper, we present GreenMind-Medium-14B-R1, the Vietnamese reasoning model inspired by the finetuning strategy based on Group Relative Policy Optimization. We also leverage a high-quality Vietnamese synthesized reasoning dataset and design two reward functions to tackle the main limitations of this technique: (i) language mixing, where we explicitly detect the presence of biased language characters during the process of sampling tokens, and (ii) we leverage Sentence Transformer-based models to ensure that the generated reasoning content maintains factual correctness and does not distort the final output. Experimental results on the Vietnamese dataset from the VLSP 2023 Challenge demonstrate that our model outperforms prior works and enhances linguistic consistency in its responses. Furthermore, we extend our evaluation to SeaExam-a multilingual multiple-choice dataset, showing the effectiveness of our reasoning method compared to few-shot prompting techniques.
Authors: Hanwen Du, Bo Peng, Xia Ning
Abstract: Target-Oriented Dialogue (TOD) remains a significant challenge in the LLM era, where strategic dialogue planning is crucial for directing conversations toward specific targets. However, existing dialogue planning methods generate dialogue plans in a step-by-step sequential manner, and may suffer from compounding errors and myopic actions. To address these limitations, we introduce a novel dialogue planning framework, DiffTOD, which leverages diffusion models to enable non-sequential dialogue planning. DiffTOD formulates dialogue planning as a trajectory generation problem with conditional guidance, and leverages a diffusion language model to estimate the likelihood of the dialogue trajectory. To optimize the dialogue action strategies, DiffTOD introduces three tailored guidance mechanisms for different target types, offering flexible guidance toward diverse TOD targets at test time. Extensive experiments across three diverse TOD settings show that DiffTOD can effectively perform non-myopic lookahead exploration and optimize action strategies over a long horizon through non-sequential dialogue planning, and demonstrates strong flexibility across complex and diverse dialogue scenarios. Our code and data are accessible through https://github.com/ninglab/DiffTOD.
Authors: Hannah Cyberey, David Evans
Abstract: Large language models (LLMs) have transformed the way we access information. These models are often tuned to refuse to comply with requests that are considered harmful and to produce responses that better align with the preferences of those who control the models. To understand how this "censorship" works. We use representation engineering techniques to study open-weights safety-tuned models. We present a method for finding a refusal--compliance vector that detects and controls the level of censorship in model outputs. We also analyze recent reasoning LLMs, distilled from DeepSeek-R1, and uncover an additional dimension of censorship through "thought suppression". We show a similar approach can be used to find a vector that suppresses the model's reasoning process, allowing us to remove censorship by applying the negative multiples of this vector. Our code is publicly available at: https://github.com/hannahxchen/llm-censorship-steering
URLs: https://github.com/hannahxchen/llm-censorship-steering
Authors: Junhong Liang, Yu Zhou
Abstract: Chinese Spelling Correction (CSC) aims to detect and correct erroneous tokens in sentences. Traditional CSC focuses on equal length correction and uses pretrained language models (PLMs). While Large Language Models (LLMs) have shown remarkable success in identifying and rectifying potential errors, they often struggle with adapting to domain-specific corrections, especially when encountering terminologies in specialized domains. To address domain adaptation, we propose a \textbf{R}etrieval-\textbf{A}ugmented \textbf{I}terative \textbf{R}efinement (RAIR) framework. Our approach constructs a retrieval corpus adaptively from domain-specific training data and dictionaries, employing a fine-tuned retriever to ensure that the retriever catches the error correction pattern. We also extend equal-length into variable-length correction scenarios. Extensive experiments demonstrate that our framework outperforms current approaches in domain spelling correction and significantly improves the performance of LLMs in variable-length scenarios.
Authors: Zimu Lu, Yunqiao Yang, Houxing Ren, Haotian Hou, Han Xiao, Ke Wang, Weikang Shi, Aojun Zhou, Mingjie Zhan, Hongsheng Li
Abstract: LLM-based agents have demonstrated great potential in generating and managing code within complex codebases. In this paper, we introduce WebGen-Bench, a novel benchmark designed to measure an LLM-based agent's ability to create multi-file website codebases from scratch. It contains diverse instructions for website generation, created through the combined efforts of human annotators and GPT-4o. These instructions span three major categories and thirteen minor categories, encompassing nearly all important types of web applications. To assess the quality of the generated websites, we use GPT-4o to generate test cases targeting each functionality described in the instructions, and then manually filter, adjust, and organize them to ensure accuracy, resulting in 647 test cases. Each test case specifies an operation to be performed on the website and the expected result after the operation. To automate testing and improve reproducibility, we employ a powerful web-navigation agent to execute tests on the generated websites and determine whether the observed responses align with the expected results. We evaluate three high-performance code-agent frameworks, Bolt.diy, OpenHands, and Aider, using multiple proprietary and open-source LLMs as engines. The best-performing combination, Bolt.diy powered by DeepSeek-R1, achieves only 27.8\% accuracy on the test cases, highlighting the challenging nature of our benchmark. Additionally, we construct WebGen-Instruct, a training set consisting of 6,667 website-generation instructions. Training Qwen2.5-Coder-32B-Instruct on Bolt.diy trajectories generated from a subset of this training set achieves an accuracy of 38.2\%, surpassing the performance of the best proprietary model.
Authors: Zixiao Zhu, Hanzhang Zhou, Zijian Feng, Tianjiao Li, Chua Jia Jim Deryl, Mak Lee Onn, Gee Wah Ng, Kezhi Mao
Abstract: Prompt optimization (PO) provides a practical way to improve response quality when users lack the time or expertise to manually craft effective prompts. Existing methods typically rely on LLMs' self-generation ability to optimize prompts. However, due to limited downward compatibility, the instruction-heavy prompts generated by advanced LLMs can overwhelm lightweight inference models and degrade response quality, while also lacking interpretability due to implicit optimization. In this work, we rethink prompt optimization through the lens of explicit and interpretable design. We first identify a set of model-agnostic prompt quality merits and empirically validate their effectiveness in enhancing prompt and response quality. We then introduce MePO, a merit-guided, locally deployable prompt optimizer trained on our merit-guided prompt preference dataset generated by a lightweight LLM. MePO avoids online optimization, reduces privacy concerns, and, by learning clear, interpretable merits, generalizes effectively to both large-scale and lightweight inference models. Experiments demonstrate that MePO achieves better results across diverse tasks and model types, offering a scalable and robust solution for real-world deployment.The code, model and dataset can be found in https://github.com/MidiyaZhu/MePO
Authors: Chunyu Ye, Yunhao Zhang, Jingyuan Sun, Chong Li, Chengqing Zong, Shaonan Wang
Abstract: Decoding language from the human brain remains a grand challenge for Brain-Computer Interfaces (BCIs). Current approaches typically rely on unimodal brain representations, neglecting the brain's inherently multimodal processing. Inspired by the brain's associative mechanisms, where viewing an image can evoke related sounds and linguistic representations, we propose a unified framework that leverages Multimodal Large Language Models (MLLMs) to align brain signals with a shared semantic space encompassing text, images, and audio. A router module dynamically selects and fuses modality-specific brain features according to the characteristics of each stimulus. Experiments on various fMRI datasets with textual, visual, and auditory stimuli demonstrate state-of-the-art performance, achieving an 8.48% improvement on the most commonly used benchmark. We further extend our framework to EEG and MEG data, demonstrating flexibility and robustness across varying temporal and spatial resolutions. To our knowledge, this is the first unified BCI architecture capable of robustly decoding multimodal brain activity across diverse brain signals and stimulus types, offering a flexible solution for real-world applications.
Authors: Hiram Ring
Abstract: Existing datasets available for crosslinguistic investigations have tended to focus on large amounts of data for a small group of languages or a small amount of data for a large number of languages. This means that claims based on these datasets are limited in what they reveal about universal properties of the human language faculty. While this has begun to change through the efforts of projects seeking to develop tagged corpora for a large number of languages, such efforts are still constrained by limits on resources. The current paper reports on a large tagged parallel dataset which has been developed to partially address this issue. The taggedPBC contains POS-tagged parallel text data from more than 1,940 languages, representing 155 language families and 78 isolates, dwarfing previously available resources. The accuracy of particular tags in this dataset is shown to correlate well with both existing SOTA taggers for high-resource languages (SpaCy, Trankit) as well as hand-tagged corpora (Universal Dependencies Treebanks). Additionally, a novel measure derived from this dataset, the N1 ratio, correlates with expert determinations of intransitive word order in three typological databases (WALS, Grambank, Autotyp) such that a Gaussian Naive Bayes classifier trained on this feature can accurately identify basic intransitive word order for languages not in those databases. While much work is still needed to expand and develop this dataset, the taggedPBC is an important step to enable corpus-based crosslinguistic investigations, and is made available for research and collaboration via GitHub.
Authors: Aliakbar Nafar, Kristen Brent Venable, Zijun Cui, Parisa Kordjamshidi
Abstract: In this work, we evaluate the potential of Large Language Models (LLMs) in building Bayesian Networks (BNs) by approximating domain expert priors. LLMs have demonstrated potential as factual knowledge bases; however, their capability to generate probabilistic knowledge about real-world events remains understudied. We explore utilizing the probabilistic knowledge inherent in LLMs to derive probability estimates for statements regarding events and their relationships within a BN. Using LLMs in this context allows for the parameterization of BNs, enabling probabilistic modeling within specific domains. Our experiments on eighty publicly available Bayesian Networks, from healthcare to finance, demonstrate that querying LLMs about the conditional probabilities of events provides meaningful results when compared to baselines, including random and uniform distributions, as well as approaches based on next-token generation probabilities. We explore how these LLM-derived distributions can serve as expert priors to refine distributions extracted from data, especially when data is scarce. Overall, this work introduces a promising strategy for automatically constructing Bayesian Networks by combining probabilistic knowledge extracted from LLMs with real-world data. Additionally, we establish the first comprehensive baseline for assessing LLM performance in extracting probabilistic knowledge.
Authors: Jialong Wu, Baixuan Li, Runnan Fang, Wenbiao Yin, Liwen Zhang, Zhengwei Tao, Dingchu Zhang, Zekun Xi, Gang Fu, Yong Jiang, Pengjun Xie, Fei Huang, Jingren Zhou
Abstract: Addressing intricate real-world problems necessitates in-depth information seeking and multi-step reasoning. Recent progress in agentic systems, exemplified by Deep Research, underscores the potential for autonomous multi-step research. In this work, we present a cohesive paradigm for building end-to-end agentic information seeking agents from a data-centric and training-stage perspective. Our approach consists of four key stages: (1) browsing data construction, (2) trajectories sampling, (3) supervised fine-tuning for effective cold start, and (4) reinforcement learning for enhanced generalisation. We instantiate this framework in a web agent based on the ReAct, WebDancer. Empirical evaluations on the challenging information seeking benchmarks, GAIA and WebWalkerQA, demonstrate the strong performance of WebDancer, achieving considerable results and highlighting the efficacy of our training paradigm. Further analysis of agent training provides valuable insights and actionable, systematic pathways for developing more capable agentic models. The codes and demo will be released in https://github.com/Alibaba-NLP/WebAgent.
Authors: Zikun Ye, Hema Yoganarasimhan
Abstract: Large Language Models (LLMs) are increasingly used in systems that retrieve and summarize content from multiple sources, such as search engines and AI assistants. While these models enhance user experience by generating coherent summaries, they obscure the contributions of original content creators, raising concerns about credit attribution and compensation. We address the challenge of valuing individual documents used in LLM-generated summaries. We propose using Shapley values, a game-theoretic method that allocates credit based on each document's marginal contribution. Although theoretically appealing, Shapley values are expensive to compute at scale. We therefore propose Cluster Shapley, an efficient approximation algorithm that leverages semantic similarity between documents. By clustering documents using LLM-based embeddings and computing Shapley values at the cluster level, our method significantly reduces computation while maintaining attribution quality. We demonstrate our approach to a summarization task using Amazon product reviews. Cluster Shapley significantly reduces computational complexity while maintaining high accuracy, outperforming baseline methods such as Monte Carlo sampling and Kernel SHAP with a better efficient frontier. Our approach is agnostic to the exact LLM used, the summarization process used, and the evaluation procedure, which makes it broadly applicable to a variety of summarization settings.
Authors: Qihui Fan, Wenbo Li, Enfu Nan, Yixiao Chen, Lei Lu, Pu Zhao, Yanzhi Wang
Abstract: The growing popularity of social deduction games has created an increasing need for intelligent frameworks where humans can collaborate with AI agents, particularly in post-pandemic contexts with heightened psychological and social pressures. Social deduction games like Werewolf, traditionally played through verbal communication, present an ideal application for Large Language Models (LLMs) given their advanced reasoning and conversational capabilities. Prior studies have shown that LLMs can outperform humans in Werewolf games, but their reliance on external modules introduces latency that left their contribution in academic domain only, and omit such game should be user-facing. We propose \textbf{Verbal Werewolf}, a novel LLM-based Werewolf game system that optimizes two parallel pipelines: gameplay powered by state-of-the-art LLMs and a fine-tuned Text-to-Speech (TTS) module that brings text output to life. Our system operates in near real-time without external decision-making modules, leveraging the enhanced reasoning capabilities of modern LLMs like DeepSeek V3 to create a more engaging and anthropomorphic gaming experience that significantly improves user engagement compared to existing text-only frameworks.
Authors: Mohammad Javad Ranjbar Kalahroodi, Amirhossein Sheikholselami, Sepehr Karimi, Sepideh Ranjbar Kalahroodi, Heshaam Faili, Azadeh Shakery
Abstract: Large Language Models (LLMs) have achieved remarkable performance on a wide range of Natural Language Processing (NLP) benchmarks, often surpassing human-level accuracy. However, their reliability in high-stakes domains such as medicine, particularly in low-resource languages, remains underexplored. In this work, we introduce PersianMedQA, a large-scale dataset of 20,785 expert-validated multiple-choice Persian medical questions from 14 years of Iranian national medical exams, spanning 23 medical specialties and designed to evaluate LLMs in both Persian and English. We benchmark 40 state-of-the-art models, including general-purpose, Persian fine-tuned, and medical LLMs, in zero-shot and chain-of-thought (CoT) settings. Our results show that closed-source general models (e.g., GPT-4.1) consistently outperform all other categories, achieving 83.09% accuracy in Persian and 80.7% in English, while Persian fine-tuned models such as Dorna underperform significantly (e.g., 34.9% in Persian), often struggling with both instruction-following and domain reasoning. We also analyze the impact of translation, showing that while English performance is generally higher, 3-10% of questions can only be answered correctly in Persian due to cultural and clinical contextual cues that are lost in translation. Finally, we demonstrate that model size alone is insufficient for robust performance without strong domain or language adaptation. PersianMedQA provides a foundation for evaluating bilingual and culturally grounded medical reasoning in LLMs. The PersianMedQA dataset is available: https://huggingface.co/datasets/MohammadJRanjbar/PersianMedQA .
URLs: https://huggingface.co/datasets/MohammadJRanjbar/PersianMedQA
Authors: Yongkang Xiao, Rui Zhang
Abstract: Multimodal knowledge graphs (MMKGs) enrich traditional knowledge graphs (KGs) by incorporating diverse modalities such as images and text. multimodal knowledge graph completion (MMKGC) seeks to exploit these heterogeneous signals to infer missing facts, thereby mitigating the intrinsic incompleteness of MMKGs. Existing MMKGC methods typically leverage only the information contained in the MMKGs under the closed-world assumption and adopt discriminative training objectives, which limits their reasoning capacity during completion. Recent large language models (LLMs), empowered by massive parameter scales and pretraining on vast corpora, have demonstrated strong reasoning abilities across various tasks. However, their potential in MMKGC remains largely unexplored. To bridge this gap, we propose HERGC, a flexible Heterogeneous Experts Representation and Generative Completion framework for MMKGs. HERGC first deploys a Heterogeneous Experts Representation Retriever that enriches and fuses multimodal information and retrieves a compact candidate set for each incomplete triple. It then uses a Generative LLM Predictor, implemented via either in-context learning or lightweight fine-tuning, to accurately identify the correct answer from these candidates. Extensive experiments on three standard MMKG benchmarks demonstrate HERGC's effectiveness and robustness, achieving superior performance over existing methods.
Authors: Dren Fazlija, Arkadij Orlov, Sandipan Sikdar
Abstract: Large language models (LLMs) are increasingly becoming valuable to corporate data management due to their ability to process text from various document formats and facilitate user interactions through natural language queries. However, LLMs must consider the sensitivity of information when communicating with employees, especially given access restrictions. Simple filtering based on user clearance levels can pose both performance and privacy challenges. To address this, we propose the concept of sensitivity awareness (SA), which enables LLMs to adhere to predefined access rights rules. In addition, we developed a benchmarking environment called ACCESS DENIED INC to evaluate SA. Our experimental findings reveal significant variations in model behavior, particularly in managing unauthorized data requests while effectively addressing legitimate queries. This work establishes a foundation for benchmarking sensitivity-aware language models and provides insights to enhance privacy-centric AI systems in corporate environments.
Authors: Lo Pang-Yun Ting, Chengshuai Zhao, Yu-Hua Zeng, Yuan Jee Lim, Kun-Ta Chuang, Huan Liu
Abstract: Clinical note generation aims to produce free-text summaries of a patient's condition and diagnostic process, with discharge instructions being a representative long-form example. While recent LLM-based methods pre-trained on general clinical corpora show promise in clinical text generation, they fall short in producing long-form notes from limited patient information. In this paper, we propose ReinRAG, a reinforced reasoning augmented generation (RAG) for long-form discharge instructions based on pre-admission information. ReinRAG retrieves reasoning paths from a medical knowledge graph to provide explicit semantic guidance to the LLM. To bridge the information gap, we propose group-based retriever optimization (GRO) which improves retrieval quality with group-normalized rewards, encouraging reasoning leaps for deeper inference by the LLM. Comprehensive experiments on the real-world dataset show that ReinRAG outperforms baselines in both clinical efficacy and natural language generation metrics. Further analysis reveals that ReinRAG fills semantic gaps in sparse input scenarios, and retrieved reasoning paths help LLMs avoid clinical misinterpretation by focusing on key evidence and following coherent reasoning.
Authors: Jash Rajesh Parekh, Pengcheng Jiang, Jiawei Han
Abstract: Retrieval-Augmented Generation (RAG) systems fail at complex multi-hop reasoning because they rely on large language models to implicitly connect information from unstructured document collections. This fundamental limitation stems from treating retrieved passages as independent context rather than recognizing the intricate relationships that enable coherent reasoning chains. We introduce SARG (Structure-Augmented Reasoning Generation), a post-retrieval framework that transforms traditional RAG pipelines by materializing explicit reasoning structures. SARG extracts {cause, relation, effect} triples from retrieved documents, constructs domain-adaptive graphs, and performs multi-hop traversal to discover reasoning chains that bridge query concepts to answers. Unlike existing approaches that modify retrieval mechanisms, SARG operates as a plug-and-play reasoning layer compatible with any RAG system. Extensive evaluation across diverse domains: general QA, biomedical literature, and financial analysis demonstrates that SARG achieves substantial improvements over state-of-the-art RAG baselines. Crucially, SARG also provides full reasoning traceability through explicit inference chains, addressing the critical interpretability gap in current RAG systems. Our results establish that explicit structural reasoning is not merely beneficial but essential for reliable complex question answering, offering a solution to RAG's implicit reasoning bottleneck.
Authors: Valentin Six, Evan Dufraisse, Ga\"el de Chalendar
Abstract: Large Language Models (LLMs) excel at many Natural Language Processing (NLP) tasks, but struggle with multi-hop reasoning and factual consistency, limiting their effectiveness on knowledge-intensive tasks like complex question answering (QA). Linking Knowledge Graphs (KG) and LLMs has shown promising results, but LLMs generally lack the ability to reason efficiently over graph-structured information. To address this challenge, we introduce DAGR, a retrieval method that leverages both complex questions and their decomposition in subquestions to extract relevant, linked textual subgraphs. DAGR first breaks down complex queries, retrieves subgraphs guided by a weighted similarity function over both the original and decomposed queries, and creates a question-specific knowledge graph to guide answer generation. The resulting Graph-RAG pipeline is suited to handle complex multi-hop questions and effectively reason over graph-structured data. We evaluate DAGR on standard multi-hop QA benchmarks and show that it achieves comparable or superior performance to competitive existing methods, using smaller models and fewer LLM calls.
Authors: Mikhail Menschikov, Dmitry Evseev, Victoria Dochkina, Ruslan Kostoev, Ilia Perepechkin, Petr Anokhin, Evgeny Burnaev, Nikita Semenov
Abstract: Personalizing language models by effectively incorporating user interaction history remains a central challenge in the development of adaptive AI systems. While large language models (LLMs) combined with Retrieval-Augmented Generation (RAG) have improved factual accuracy, they often lack structured memory and fail to scale in complex, long-term interactions. To address this, we propose a flexible external memory framework based on knowledge graphs, automatically constructed and updated by the LLM itself, and capable of encoding information in multiple formats-including nodes, triplets, higher-order propositions, and episodic traces. Building upon the AriGraph architecture, we introduce a novel hybrid graph design that supports both standard edges and two types of hyperedges, enabling rich and dynamic semantic and temporal representations. Our framework also supports diverse retrieval mechanisms, including A*, water-circle propagation, beam search, and hybrid methods, making it adaptable to different datasets and LLM capacities. We evaluate our system on three benchmarks-TriviaQA, HotpotQA, and DiaASQ-demonstrating that different memory and retrieval configurations yield optimal performance depending on the task. Additionally, we extend the DiaASQ benchmark with temporal annotations and internally contradictory statements, showing that our system remains robust and effective in managing temporal dependencies and context-aware reasoning.
Authors: Deepon Halder, Thanmay Jayakumar, Raj Dabre
Abstract: Large language models (LLMs), despite their ability to perform few-shot machine translation (MT), often lag behind dedicated MT systems trained on parallel corpora, which are crucial for high quality machine translation (MT). However, parallel corpora are often scarce or non-existent for low-resource languages. In this paper, we propose CycleDistill, a bootstrapping approach leveraging LLMs and few-shot translation to obtain high-quality MT systems. CycleDistill involves iteratively generating synthetic parallel corpora from monolingual corpora via zero- or few-shot MT, which is then used to fine-tune the model that was used for generating said data for MT. CycleDistill does not need parallel corpora beyond 1 to 4 few-shot examples, and in our experiments focusing on three Indian languages, by relying solely on monolingual corpora, it can achieve high-quality machine translation, improving upon a few-shot baseline model by over 20-30 chrF points on average in the first iteration. We also study the effect of leveraging softmax activations during the distillation process and observe mild improvements in translation quality.
Authors: Chris Madge, Maris Camilleri, Paloma Carretero Garcia, Vanja Karan, Juexi Shao, Prashant Jayannavar, Julian Hough, Benjamin Roth, Massimo Poesio
Abstract: We introduce the Minecraft Dialogue Corpus with Reference (MDC-R). MDC-R is a new language resource that supplements the original Minecraft Dialogue Corpus (MDC) with expert annotations of anaphoric and deictic reference. MDC's task-orientated, multi-turn, situated dialogue in a dynamic environment has motivated multiple annotation efforts, owing to the interesting linguistic phenomena that this setting gives rise to. We believe it can serve as a valuable resource when annotated with reference, too. Here, we discuss our method of annotation and the resulting corpus, and provide both a quantitative and a qualitative analysis of the data. Furthermore, we carry out a short experiment demonstrating the usefulness of our corpus for referring expression comprehension.
Authors: Seungjun Yi, Joakim Nguyen, Huimin Xu, Terence Lim, Andrew Well, Mia Markey, Ying Ding
Abstract: Congenital heart disease (CHD) presents complex, lifelong challenges often underrepresented in traditional clinical metrics. While unstructured narratives offer rich insights into patient and caregiver experiences, manual thematic analysis (TA) remains labor-intensive and unscalable. We propose a fully automated large language model (LLM) pipeline that performs end-to-end TA on clinical narratives, which eliminates the need for manual coding or full transcript review. Our system employs a novel multi-agent framework, where specialized LLM agents assume roles to enhance theme quality and alignment with human analysis. To further improve thematic relevance, we optionally integrate reinforcement learning from human feedback (RLHF). This supports scalable, patient-centered analysis of large qualitative datasets and allows LLMs to be fine-tuned for specific clinical contexts.
Authors: Anselm R. Strohmaier, Wim Van Dooren, Kathrin Se{\ss}ler, Brian Greer, Lieven Verschaffel
Abstract: The progress of Large Language Models (LLMs) like ChatGPT raises the question of how they can be integrated into education. One hope is that they can support mathematics learning, including word-problem solving. Since LLMs can handle textual input with ease, they appear well-suited for solving mathematical word problems. Yet their real competence, whether they can make sense of the real-world context, and the implications for classrooms remain unclear. We conducted a scoping review from a mathematics-education perspective, including three parts: a technical overview, a systematic review of word problems used in research, and a state-of-the-art empirical evaluation of LLMs on mathematical word problems. First, in the technical overview, we contrast the conceptualization of word problems and their solution processes between LLMs and students. In computer-science research this is typically labeled mathematical reasoning, a term that does not align with usage in mathematics education. Second, our literature review of 213 studies shows that the most popular word-problem corpora are dominated by s-problems, which do not require a consideration of realities of their real-world context. Finally, our evaluation of GPT-3.5-turbo, GPT-4o-mini, GPT-4.1, o3, and GPT-5 on 287 word problems shows that most recent LLMs solve these s-problems with near-perfect accuracy, including a perfect score on 20 problems from PISA. LLMs still showed weaknesses in tackling problems where the real-world context is problematic or non-sensical. In sum, we argue based on all three aspects that LLMs have mastered a superficial solution process but do not make sense of word problems, which potentially limits their value as instructional tools in mathematics classrooms.
Authors: Guanzhong Pan, Mei Tan, Hyunji Nam, Luc\'ia Langlois, James Malamut, Liliana Deonizio, Dorottya Demszky
Abstract: We introduce EduCoder, a domain-specialized tool designed to support utterance-level annotation of educational dialogue. While general-purpose text annotation tools for NLP and qualitative research abound, few address the complexities of coding education dialogue transcripts -- with diverse teacher-student and peer interactions. Common challenges include defining codebooks for complex pedagogical features, supporting both open-ended and categorical coding, and contextualizing utterances with external features, such as the lesson's purpose and the pedagogical value of the instruction. EduCoder is designed to address these challenges by providing a platform for researchers and domain experts to collaboratively define complex codebooks based on observed data. It incorporates both categorical and open-ended annotation types along with contextual materials. Additionally, it offers a side-by-side comparison of multiple annotators' responses, allowing comparison and calibration of annotations with others to improve data reliability. The system is open-source, with a demo video available.
Authors: Changxin Tian, Jiapeng Wang, Qian Zhao, Kunlong Chen, Jia Liu, Ziqi Liu, Jiaxin Mao, Wayne Xin Zhao, Zhiqiang Zhang, Jun Zhou
Abstract: Recent advances in learning rate (LR) scheduling have demonstrated the effectiveness of decay-free approaches that eliminate the traditional decay phase while maintaining competitive performance. Model merging techniques have emerged as particularly promising solutions in this domain. We present Warmup-Stable and Merge (WSM), a general framework that establishes a formal connection between learning rate decay and model merging. WSM provides a unified theoretical foundation for emulating various decay strategies-including cosine decay, linear decay and inverse square root decay-as principled model averaging schemes, while remaining fully compatible with diverse optimization methods. Through extensive experiments, we identify merge duration-the training window for checkpoint aggregation-as the most critical factor influencing model performance, surpassing the importance of both checkpoint interval and merge quantity. Our framework consistently outperforms the widely-adopted Warmup-Stable-Decay (WSD) approach across multiple benchmarks, achieving significant improvements of +3.5% on MATH, +2.9% on HumanEval, and +5.5% on MMLU-Pro. The performance advantages extend to supervised fine-tuning scenarios, highlighting WSM's potential for long-term model refinement.
Authors: Changxin Tian, Kunlong Chen, Jia Liu, Ziqi Liu, Zhiqiang Zhang, Jun Zhou
Abstract: Mixture-of-Experts (MoE) has become a dominant architecture for scaling Large Language Models (LLMs) efficiently by decoupling total parameters from computational cost. However, this decoupling creates a critical challenge: predicting the model capacity of a given MoE configurations (e.g., expert activation ratio and granularity) remains an unresolved problem. To address this gap, we introduce Efficiency Leverage (EL), a metric quantifying the computational advantage of an MoE model over a dense equivalent. We conduct a large-scale empirical study, training over 300 models up to 28B parameters, to systematically investigate the relationship between MoE architectural configurations and EL. Our findings reveal that EL is primarily driven by the expert activation ratio and the total compute budget, both following predictable power laws, while expert granularity acts as a non-linear modulator with a clear optimal range. We integrate these discoveries into a unified scaling law that accurately predicts the EL of an MoE architecture based on its configuration. To validate our derived scaling laws, we designed and trained Ling-mini-beta, a pilot model for Ling-2.0 series with only 0.85B active parameters, alongside a 6.1B dense model for comparison. When trained on an identical 1T high-quality token dataset, Ling-mini-beta matched the performance of the 6.1B dense model while consuming over 7x fewer computational resources, thereby confirming the accuracy of our scaling laws. This work provides a principled and empirically-grounded foundation for the scaling of efficient MoE models.
Authors: Daniil Gurgurov, Katharina Trinley, Yusser Al Ghussin, Tanja Baeumel, Josef van Genabith, Simon Ostermann
Abstract: Large language models (LLMs) exhibit strong multilingual abilities, yet the neural mechanisms behind language-specific processing remain unclear. We analyze language-specific neurons in Llama-3.1-8B, Mistral-Nemo-12B, and Aya-Expanse-8B & 32B across 21 typologically diverse languages, identifying neurons that control language behavior. Using the Language Activation Probability Entropy (LAPE) method, we show that these neurons cluster in deeper layers, with non-Latin scripts showing greater specialization. Related languages share overlapping neurons, reflecting internal representations of linguistic proximity. Through language arithmetics, i.e. systematic activation addition and multiplication, we steer models to deactivate unwanted languages and activate desired ones, outperforming simpler replacement approaches. These interventions effectively guide behavior across five multilingual tasks: language forcing, translation, QA, comprehension, and NLI. Manipulation is more successful for high-resource languages, while typological similarity improves effectiveness. We also demonstrate that cross-lingual neuron steering enhances downstream performance and reveal internal "fallback" mechanisms for language selection when neurons are progressively deactivated. Our code is made publicly available at https://github.com/d-gurgurov/Language-Neurons-Manipulation.
URLs: https://github.com/d-gurgurov/Language-Neurons-Manipulation.
Authors: Qixuan Hu, Xumou Zhang, Jinman Kim, Florence Bourgeois, Adam G. Dunn
Abstract: Objectives: With accurate estimates of expected safety results, clinical trials could be better designed and monitored. We evaluated methods for predicting serious adverse event (SAE) results in clinical trials using information only from their registrations prior to the trial. Material and Methods: We analyzed 22,107 two-arm parallel interventional clinical trials from ClinicalTrials.gov with structured summary results. Two prediction models were developed: a classifier predicting whether a greater proportion of participants in an experimental arm would have SAEs (area under the receiver operating characteristic curve; AUC) compared to the control arm, and a regression model to predict the proportion of participants with SAEs in the control arms (root mean squared error; RMSE). A transfer learning approach using pretrained language models (e.g., ClinicalT5, BioBERT) was used for feature extraction, combined with a downstream model for prediction. To maintain semantic representation in long trial texts exceeding localized language model input limits, a sliding window method was developed for embedding extraction. Results: The best model (ClinicalT5+Transformer+MLP) had 77.6% AUC when predicting which trial arm had a higher proportion of SAEs. When predicting SAE proportion in the control arm, the same model achieved RMSE of 18.6%. The sliding window approach consistently outperformed direct comparisons. Across 12 classifiers, the average absolute AUC increase was 2.00%, and absolute RMSE reduction was 1.58% across 12 regressors. Discussion: Summary results data from ClinicalTrials.gov remains underutilized. Predicted results of publicly reported trials provides an opportunity to identify discrepancies between expected and reported safety results.
Authors: Xiao Zhang, Johan bos
Abstract: Neural semantic parsers have shown good overall performance for a variety of linguistic phenomena, reaching semantic matching scores of more than 90%. But how do such parsers perform on strongly context-sensitive phenomena, where large pieces of semantic information need to be duplicated to form a meaningful semantic representation? A case in point is English verb phrase ellipsis, a construct where entire verb phrases can be abbreviated by a single auxiliary verb. Are the otherwise known as powerful semantic parsers able to deal with ellipsis or aren't they? We constructed a corpus of 120 cases of ellipsis with their fully resolved meaning representation and used this as a challenge set for a large battery of neural semantic parsers. Although these parsers performed very well on the standard test set, they failed in the instances with ellipsis. Data augmentation helped improve the parsing results. The reason for the difficulty of parsing elided phrases is not that copying semantic material is hard, but that usually occur in linguistically complicated contexts causing most of the parsing errors.
Authors: Minghao Guo, Xi Zhu, Jingyuan Huang, Kai Mei, Yongfeng Zhang
Abstract: Graph Neural Networks (GNNs) have achieved remarkable success in graph-based learning by propagating information among neighbor nodes via predefined aggregation mechanisms. However, such fixed schemes often suffer from two key limitations. First, they cannot handle the imbalance in node informativeness -- some nodes are rich in information, while others remain sparse. Second, predefined message passing primarily leverages local structural similarity while ignoring global semantic relationships across the graph, limiting the model's ability to capture distant but relevant information. We propose Retrieval-augmented Graph Agentic Network (ReaGAN), an agent-based framework that empowers each node with autonomous, node-level decision-making. Each node acts as an agent that independently plans its next action based on its internal memory, enabling node-level planning and adaptive message propagation. Additionally, retrieval-augmented generation (RAG) allows nodes to access semantically relevant content and build global relationships in the graph. ReaGAN achieves competitive performance under few-shot in-context settings using a frozen LLM backbone without fine-tuning, showcasing the potential of agentic planning and local-global retrieval in graph learning.
Authors: Muhammed Saeed, Shaina Raza, Ashmal Vayani, Muhammad Abdul-Mageed, Ali Emami, Shady Shehata
Abstract: Research on bias in Text-to-Image (T2I) models has primarily focused on demographic representation and stereotypical attributes, overlooking a fundamental question: how does grammatical gender influence visual representation across languages? We introduce a cross-linguistic benchmark examining words where grammatical gender contradicts stereotypical gender associations (e.g., ``une sentinelle'' - grammatically feminine in French but referring to the stereotypically masculine concept ``guard''). Our dataset spans five gendered languages (French, Spanish, German, Italian, Russian) and two gender-neutral control languages (English, Chinese), comprising 800 unique prompts that generated 28,800 images across three state-of-the-art T2I models. Our analysis reveals that grammatical gender dramatically influences image generation: masculine grammatical markers increase male representation to 73\% on average (compared to 22\% with gender-neutral English), while feminine grammatical markers increase female representation to 38\% (compared to 28\% in English). These effects vary systematically by language resource availability and model architecture, with high-resource languages showing stronger effects. Our findings establish that language structure itself, not just content, shapes AI-generated visual outputs, introducing a new dimension for understanding bias and fairness in multilingual, multimodal systems.
Authors: Pablo J. Diego-Sim\'on, Emmanuel Chemla, Jean-R\'emi King, Yair Lakretz
Abstract: The syntactic structures of sentences can be readily read-out from the activations of large language models (LLMs). However, the ``structural probes'' that have been developed to reveal this phenomenon are typically evaluated on an indiscriminate set of sentences. Consequently, it remains unclear whether structural and/or statistical factors systematically affect these syntactic representations. To address this issue, we conduct an in-depth analysis of structural probes on three controlled benchmarks. Our results are three-fold. First, structural probes are biased by a superficial property: the closer two words are in a sentence, the more likely structural probes will consider them as syntactically linked. Second, structural probes are challenged by linguistic properties: they poorly represent deep syntactic structures, and get interfered by interacting nouns or ungrammatical verb forms. Third, structural probes do not appear to be affected by the predictability of individual words. Overall, this work sheds light on the current challenges faced by structural probes. Providing a benchmark made of controlled stimuli to better evaluate their performance.
Authors: Yajie Luo, Yihong Wu, Muzhi Li, Fengran Mo, Jia Ao Sun, Xinyu Wang, Liheng Ma, Yingxue Zhang, Jian-Yun Nie
Abstract: Some Question Answering (QA) systems rely on knowledge bases (KBs) to provide accurate answers. Entity Linking (EL) plays a critical role in linking natural language mentions to KB entries. However, most existing EL methods are designed for long contexts and do not perform well on short, ambiguous user questions in QA tasks. We propose an entity linking agent for QA, based on a Large Language Model that simulates human cognitive workflows. The agent actively identifies entity mentions, retrieves candidate entities, and makes decision. To verify the effectiveness of our agent, we conduct two experiments: tool-based entity linking and QA task evaluation. The results confirm the robustness and effectiveness of our agent.
Authors: Jun Liu, Zhenglun Kong, Changdi Yang, Fan Yang, Tianqi Li, Peiyan Dong, Joannah Nanjekye, Hao Tang, Geng Yuan, Wei Niu, Wenbin Zhang, Pu Zhao, Xue Lin, Dong Huang, Yanzhi Wang
Abstract: Multi-agent large language model (LLM) systems have shown strong potential in complex reasoning and collaborative decision-making tasks. However, most existing coordination schemes rely on static or full-context routing strategies, which lead to excessive token consumption, redundant memory exposure, and limited adaptability across interaction rounds. We introduce RCR-Router, a modular and role-aware context routing framework designed to enable efficient, adaptive collaboration in multi-agent LLMs. To our knowledge, this is the first routing approach that dynamically selects semantically relevant memory subsets for each agent based on its role and task stage, while adhering to a strict token budget. A lightweight scoring policy guides memory selection, and agent outputs are iteratively integrated into a shared memory store to facilitate progressive context refinement. To better evaluate model behavior, we further propose an Answer Quality Score metric that captures LLM-generated explanations beyond standard QA accuracy. Experiments on three multi-hop QA benchmarks -- HotPotQA, MuSiQue, and 2WikiMultihop -- demonstrate that RCR-Router reduces token usage (up to 30%) while improving or maintaining answer quality. These results highlight the importance of structured memory routing and output-aware evaluation in advancing scalable multi-agent LLM systems.
Authors: Aditya Kishore, Gaurav Kumar, Jasabanta Patro
Abstract: The growing rate of multimodal misinformation, where claims are supported by both text and images, poses significant challenges to fact-checking systems that rely primarily on textual evidence. In this work, we have proposed a unified framework for fine-grained multimodal fact verification called "MultiCheck", designed to reason over structured textual and visual signals. Our architecture combines dedicated encoders for text and images with a fusion module that captures cross-modal relationships using element-wise interactions. A classification head then predicts the veracity of a claim, supported by a contrastive learning objective that encourages semantic alignment between claim-evidence pairs in a shared latent space. We evaluate our approach on the Factify 2 dataset, achieving a weighted F1 score of 0.84, substantially outperforming the baseline. These results highlight the effectiveness of explicit multimodal reasoning and demonstrate the potential of our approach for scalable and interpretable fact-checking in complex, real-world scenarios.
Authors: Yilin Xiao, Chuang Zhou, Qinggang Zhang, Su Dong, Shengyuan Chen, Xiao Huang
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, yet exhibit critical limitations in knowledge-intensive tasks, often generating hallucinations when faced with questions requiring specialized expertise. While retrieval-augmented generation (RAG) mitigates this by integrating external knowledge, it struggles with complex reasoning scenarios due to its reliance on direct semantic retrieval and lack of structured logical organization. Inspired by Cartesian principles from \textit{Discours de la m\'ethode}, this paper introduces Logic-Augmented Generation (LAG), a novel paradigm that reframes knowledge augmentation through systematic question decomposition and dependency-aware reasoning. Specifically, LAG first decomposes complex questions into atomic sub-questions ordered by logical dependencies. It then resolves these sequentially, using prior answers to guide context retrieval for subsequent sub-questions, ensuring stepwise grounding in logical chain. To prevent error propagation, LAG incorporates a logical termination mechanism that halts inference upon encountering unanswerable sub-questions and reduces wasted computation on excessive reasoning. Finally, it synthesizes all sub-resolutions to generate verified responses. Experiments on four benchmark datasets demonstrate that LAG significantly enhances reasoning robustness, reduces hallucination, and aligns LLM problem-solving with human cognition, offering a principled alternative to existing RAG systems.
Authors: Shaoxiong Zhan, Yanlin Lai, Ziyu Lu, Dahua Lin, Ziqing Yang, Fei Tan
Abstract: Large language models have achieved substantial progress in mathematical reasoning, yet their advancement is limited by the scarcity of high-quality, high-difficulty training data. Existing synthesis methods largely rely on transforming human-written templates, limiting both diversity and scalability. We propose MathSmith, a novel framework for synthesizing challenging mathematical problems to enhance LLM reasoning. Rather than modifying existing problems, MathSmith constructs new ones from scratch by randomly sampling concept-explanation pairs from PlanetMath, ensuring data independence and avoiding contamination. To increase difficulty, we design nine predefined strategies as soft constraints during rationales. We further adopts reinforcement learning to jointly optimize structural validity, reasoning complexity, and answer consistency. The length of the reasoning trace generated under autoregressive prompting is used to reflect cognitive complexity, encouraging the creation of more demanding problems aligned with long-chain-of-thought reasoning. Experiments across five benchmarks, categorized as easy & medium (GSM8K, MATH-500) and hard (AIME2024, AIME2025, OlympiadBench), show that MathSmith consistently outperforms existing baselines under both short and long CoT settings. Additionally, a weakness-focused variant generation module enables targeted improvement on specific concepts. Overall, MathSmith exhibits strong scalability, generalization, and transferability, highlighting the promise of high-difficulty synthetic data in advancing LLM reasoning capabilities.
Authors: Kara Kedrick, Ekaterina Levitskaya, Russell J. Funk
Abstract: A growing stream of research finds that scientific contributions are evaluated differently depending on the gender of the author. In this article, we consider whether gender differences in writing styles - how men and women communicate their work - may contribute to these observed gender gaps. We ground our investigation in a framework for characterizing the linguistic style of written text, with two sets of features - informational (i.e., features that emphasize facts) and involved (i.e., features that emphasize relationships). Using a large sample of academic papers and patents, we find significant differences in writing style by gender, with women using more involved features in their writing. Papers and patents with more involved features also tend to be cited more by women. Our findings suggest that scientific text is not devoid of personal character, which could contribute to bias in evaluation, thereby compromising the norm of universalism as a foundational principle of science.
Authors: Prateek Yadav, Leshem Choshen, Colin Raffel, Mohit Bansal
Abstract: Parameter-efficient fine-tuning (PEFT) techniques make it possible to efficiently adapt a language model to create "expert" models that specialize to new tasks or domains. Recent techniques in model merging and compositional generalization leverage these expert models by dynamically composing modules to improve zero/few-shot generalization. Despite the efficiency of PEFT methods, the size of expert models can make it onerous to retrieve expert models per query over high-latency networks like the Internet or serve multiple experts on a single GPU. To address these issues, we present ComPEFT, a novel method for compressing fine-tuning residuals (task vectors) of PEFT based models. ComPEFT employs sparsification and ternary quantization to reduce the size of the PEFT module without performing any additional retraining while preserving or enhancing model performance. In extensive evaluation across T5, T0, and LLaMA-based models with 200M - 65B parameters, ComPEFT achieves compression ratios of 8x - 50x. In particular, we show that ComPEFT improves with scale - stronger models exhibit higher compressibility and better performance. For example, we show that ComPEFT applied to LLaMA outperforms QLoRA by 4.16% on MMLU with a storage size reduction of up to 26x. In addition, we show that the compressed experts produced by ComPEFT maintain few-shot compositional generalization capabilities, facilitate efficient communication and computation, and exhibit enhanced performance when merged. Lastly, we provide an analysis of different method components, compare it with other PEFT methods, and test ComPEFT's efficacy for compressing the residual of full-finetuning. Our code is available at https://github.com/prateeky2806/compeft.
Authors: Zheng Liu, Hao Liang, Bozhou Li, Wentao Xiong, Chong Chen, Conghui He, Wentao Zhang, Bin Cui
Abstract: Vision-Language Models (VLMs) have recently emerged, demonstrating remarkable vision-understanding capabilities. However, training these models requires large-scale datasets, which brings challenges related to efficiency, effectiveness, and quality of web data. In this paper, we introduce SynthVLM, a new data synthesis and curation method for generating image-caption pairs. Unlike traditional methods, where captions are generated from images, SynthVLM utilizes advanced diffusion models and high-quality captions to synthesize and select images from text captions, thereby creating precisely aligned image-text pairs. We further introduce SynthVLM-100K, a high-quality dataset consisting of 100K curated and synthesized image-caption pairs. In both model and human evaluations, SynthVLM-100K outperforms traditional real-world datasets. Leveraging this dataset, we develop a new family of multimodal large language models (MLLMs), SynthVLM-7B and SynthVLM-13B, which achieve state-of-the-art (SOTA) performance on various vision question-answering (VQA) tasks. Notably, our models outperform LLaVA across most metrics with only 18\% pretrain data. Furthermore, SynthVLM-7B and SynthVLM-13B attain SOTA performance on the MMLU benchmark, demonstrating that the high-quality SynthVLM-100K dataset preserves language abilities.
Authors: Prateek Yadav, Colin Raffel, Mohammed Muqeeth, Lucas Caccia, Haokun Liu, Tianlong Chen, Mohit Bansal, Leshem Choshen, Alessandro Sordoni
Abstract: The availability of performant pre-trained models has led to a proliferation of fine-tuned expert models that are specialized to a particular domain or task. Model MoErging methods aim to recycle expert models to create an aggregate system with improved performance or generalization. A key component of MoErging methods is the creation of a router that decides which expert model(s) to use for a particular input or application. The promise, effectiveness, and large design space of MoErging has spurred the development of many new methods over the past few years. This rapid pace of development has made it challenging to compare different MoErging methods, which are rarely compared to one another and are often validated in different experimental setups. To remedy such gaps, we present a comprehensive survey of MoErging methods that includes a novel taxonomy for cataloging key design choices and clarifying suitable applications for each method. Apart from surveying MoErging research, we inventory software tools and applications that make use of MoErging. We additionally discuss related fields of study such as model merging, multitask learning, and mixture-of-experts models. Taken as a whole, our survey provides a unified overview of existing MoErging methods and creates a solid foundation for future work in this burgeoning field.
Authors: Hao Liang, Linzhuang Sun, Minxuan Zhou, Zirong Chen, Meiyi Qiang, Mingan Lin, Tianpeng Li, Fan Yang, Zenan Zhou, Wentao Zhang
Abstract: With the rapid progress of Multimodal LLMs, evaluating their mathematical reasoning capabilities has become an increasingly important research direction. In particular, visual-textual mathematical reasoning serves as a key indicator of an MLLM's ability to comprehend and solve complex, multi-step quantitative problems. While existing benchmarks such as MathVista and MathVerse have advanced the evaluation of multimodal math proficiency, they primarily rely on digitally rendered content and fall short in capturing the complexity of real-world scenarios. To bridge this gap, we introduce MathScape, a novel benchmark focused on assessing MLLMs' reasoning ability in realistic mathematical contexts. MathScape comprises 1,369 high-quality math problems paired with human-captured real-world images, closely reflecting the challenges encountered in practical educational settings. We conduct a thorough multi-dimensional evaluation across nine leading closed-source MLLMs, three open-source MLLMs with over 20 billion parameters, and seven smaller-scale MLLMs. Our results show that even SOTA models struggle with real-world math tasks, lagging behind human performance -- highlighting critical limitations in current model capabilities. Moreover, we find that strong performance on synthetic or digitally rendered images does not guarantee similar effectiveness on real-world tasks. This underscores the necessity of MathScape in the next stage of multimodal mathematical reasoning.
Authors: Hongcheng Gao, Tianyu Pang, Chao Du, Taihang Hu, Zhijie Deng, Min Lin
Abstract: With the rapid progress of diffusion-based content generation, significant efforts are being made to unlearn harmful or copyrighted concepts from pretrained diffusion models (DMs) to prevent potential model misuse. However, it is observed that even when DMs are properly unlearned before release, malicious finetuning can compromise this process, causing DMs to relearn the unlearned concepts. This occurs partly because certain benign concepts (e.g., "skin") retained in DMs are related to the unlearned ones (e.g., "nudity"), facilitating their relearning via finetuning. To address this, we propose meta-unlearning on DMs. Intuitively, a meta-unlearned DM should behave like an unlearned DM when used as is; moreover, if the meta-unlearned DM undergoes malicious finetuning on unlearned concepts, the related benign concepts retained within it will be triggered to self-destruct, hindering the relearning of unlearned concepts. Our meta-unlearning framework is compatible with most existing unlearning methods, requiring only the addition of an easy-to-implement meta objective. We validate our approach through empirical experiments on meta-unlearning concepts from Stable Diffusion models (SD-v1-4 and SDXL), supported by extensive ablation studies. Our code is available at https://github.com/sail-sg/Meta-Unlearning.
Authors: Yuci Liang, Xinheng Lyu, Wenting Chen, Meidan Ding, Jipeng Zhang, Xiangjian He, Song Wu, Xiaohan Xing, Sen Yang, Xiyue Wang, Linlin Shen
Abstract: Recent advancements in computational pathology have produced patch-level Multi-modal Large Language Models (MLLMs), but these models are limited by their inability to analyze whole slide images (WSIs) comprehensively and their tendency to bypass crucial morphological features that pathologists rely on for diagnosis. To address these challenges, we first introduce WSI-Bench, a large-scale morphology-aware benchmark containing 180k VQA pairs from 9,850 WSIs across 30 cancer types, designed to evaluate MLLMs' understanding of morphological characteristics crucial for accurate diagnosis. Building upon this benchmark, we present WSI-LLaVA, a novel framework for gigapixel WSI understanding that employs a three-stage training approach: WSI-text alignment, feature space alignment, and task-specific instruction tuning. To better assess model performance in pathological contexts, we develop two specialized WSI metrics: WSI-Precision and WSI-Relevance. Experimental results demonstrate that WSI-LLaVA outperforms existing models across all capability dimensions, with a significant improvement in morphological analysis, establishing a clear correlation between morphological understanding and diagnostic accuracy.
Authors: Angelina Wang, Michelle Phan, Daniel E. Ho, Sanmi Koyejo
Abstract: Algorithmic fairness has conventionally adopted the mathematically convenient perspective of racial color-blindness (i.e., difference unaware treatment). However, we contend that in a range of important settings, group difference awareness matters. For example, differentiating between groups may be necessary in legal contexts (e.g., the U.S. compulsory draft applies to men but not women) and harm assessments (e.g., referring to girls as ``terrorists'' may be less harmful than referring to Muslim people as such). Thus, in contrast to most fairness work, we study fairness through the perspective of treating people differently -- when it is contextually appropriate to. We first introduce an important distinction between descriptive (fact-based), normative (value-based), and correlation (association-based) benchmarks. This distinction is significant because each category requires separate interpretation and mitigation tailored to its specific characteristics. Then, we present a benchmark suite composed of eight different scenarios for a total of 16k questions that enables us to assess difference awareness. Finally, we show results across ten models that demonstrate difference awareness is a distinct dimension to fairness where existing bias mitigation strategies may backfire.
Authors: Xuefeng Liu, Songhao Jiang, Ian Foster, Jinbo Xu, Rick Stevens
Abstract: Drug optimization has become increasingly crucial in light of fast-mutating virus strains and drug-resistant cancer cells. Nevertheless, it remains challenging as it necessitates retaining the beneficial properties of the original drug while simultaneously enhancing desired attributes beyond its scope. In this work, we aim to tackle this challenge by introducing ScaffoldGPT, a novel Generative Pretrained Transformer (GPT) designed for drug optimization based on molecular scaffolds. Our work comprises three key components: (1) A three-stage drug optimization approach that integrates pretraining, finetuning, and decoding optimization. (2) A novel two-phase incremental pre-training strategy for scaffold-based drug optimization. (3) A token-level decoding optimization strategy, Top-N, that enabling controlled, reward-guided generation using the pretrained or finetuned GPT. We demonstrate via a comprehensive evaluation on COVID and cancer benchmarks that ScaffoldGPT outperforms the competing baselines in drug optimization benchmarks, while excelling in preserving original functional scaffold and enhancing desired properties.
Authors: Lucas Torroba-Hennigen, Hunter Lang, Han Guo, Yoon Kim
Abstract: We study memory-efficient optimization of neural networks (in particular language models) with linear gradient transformations, where the gradients are linearly mapped to a lower dimensional space than the full parameter space, thus saving memory required for gradient accumulation and optimizer state persistence. The model parameters are updated by first performing an optimization step in the lower dimensional space and then going back into the original parameter space via the linear map's transpose. We show that optimizing the model in this transformed space is equivalent to reparameterizing the original model through a linear adapter that additively modifies the model parameters, and then only optimizing the adapter's parameters. When the transformation is Kronecker-factored, this establishes an equivalence between GaLore and one-sided LoRA. We show that this duality between gradient transformations and adapter-based reparameterizations unifies existing approaches to memory-efficient training and suggests new techniques for improving training efficiency and memory use.
Authors: Seyed Pouyan Mousavi Davoudi, Amin Gholami Davodi, Alireza Amiri-Margavi, Alireza Shafiee Fard, Mahdi Jafari
Abstract: We introduce a new approach in which several advanced large language models-specifically GPT-4-0125-preview, Meta-LLAMA-3-70B-Instruct, Claude-3-Opus, and Gemini-1.5-Flash-collaborate to both produce and answer intricate, doctoral-level probability problems without relying on any single "correct" reference. Rather than depending on an established ground truth, our investigation focuses on how agreement among diverse models can signal the reliability of their outputs and, by extension, reflect the overall quality of the generated questions. To measure this inter-model alignment, we apply a suite of statistical evaluations, including chi-square tests, Fleiss' Kappa coefficients, and confidence interval calculations, thereby capturing both precision in answers and clarity in question phrasing. Our analysis reveals that Claude and Gemini tend to frame questions more coherently and unambiguously, which is evidenced by their tighter confidence intervals and greater concordance with responding agents. In contrast, LLAMA exhibits wider confidence bands and a lower level of agreement, indicating more variability and reduced consistency in its question formulations. These observations support the notion that a multi-model collaborative strategy not only improves answer dependability but also offers an effective, data-driven mechanism for evaluating and refining question quality when no definitive solution exists. Ultimately, this work delivers actionable insights into enhancing AI-guided reasoning processes through coordinated interactions among heterogeneous language models.
Authors: Qiyuan Li, Haijiang Liu, Caicai Guo, Chao Gao, Deyu Chen, Meng Wang, Feng Gao, Frank van Harmelen, Jinguang Gu
Abstract: The large-scale development of large language models (LLMs) in medical contexts, such as diagnostic assistance and treatment recommendations, necessitates that these models possess accurate medical knowledge and deliver traceable decision-making processes. Clinical knowledge, encompassing the insights gained from research on the causes, prognosis, diagnosis, and treatment of diseases, has been extensively examined within real-world medical practices. Recently, there has been a notable increase in research efforts aimed at integrating this type of knowledge into LLMs, encompassing not only traditional text and multimodal data integration but also technologies such as knowledge graphs (KGs) and retrieval-augmented generation (RAG). In this paper, we review the various initiatives to embed clinical knowledge into training-based, KG-supported, and RAG-assisted LLMs. We begin by gathering reliable knowledge sources from the medical domain, including databases and datasets. Next, we evaluate implementations for integrating clinical knowledge through specialized datasets and collaborations with external knowledge sources such as KGs and relevant documentation. Furthermore, we discuss the applications of the developed medical LLMs in the industrial sector to assess the disparity between models developed in academic settings and those in industry. We conclude the survey by presenting evaluation systems applicable to relevant tasks and identifying potential challenges facing this field. In this review, we do not aim for completeness, since any ostensibly complete review would soon be outdated. Our goal is to illustrate diversity by selecting representative and accessible items from current research and industry practices, reflecting real-world situations rather than claiming completeness. Thus, we emphasize showcasing diverse approaches.
Authors: Kyle Gao, Dening Lu, Liangzhi Li, Nan Chen, Hongjie He, Linlin Xu, Jonathan Li
Abstract: The Los Angeles wildfires of January 2025 caused more than 250 billion dollars in damage and lasted for nearly an entire month before containment. Following our previous work, the Digital Twin Building, we modify and leverage the multi-agent large language model framework as well as the cloud-mapping integration to study the air quality during the Los Angeles wildfires. Recent advances in large language models have allowed for out-of-the-box automated large-scale data analysis. We use a multi-agent large language system comprised of an Instructor agent and Worker agents. Upon receiving the users' instructions, the Instructor agent retrieves the data from the cloud platform and produces instruction prompts to the Worker agents. The Worker agents then analyze the data and provide summaries. The summaries are finally input back into the Instructor agent, which then provides the final data analysis. We test this system's capability for data-based policy recommendation by assessing our Instructor-Worker LLM system's health recommendations based on air quality during the Los Angeles wildfires.
Authors: Annie Wong, Thomas B\"ack, Aske Plaat, Niki van Stein, Anna V. Kononova
Abstract: Large language models excel on static benchmarks, but their ability as self-learning agents in dynamic environments remains unclear. We evaluate three prompting strategies: self-reflection, heuristic mutation, and planning across dynamic tasks with open-source models. We find that larger models generally outperform smaller ones, but that strategic prompting can close this performance gap. Second, an overly long prompt can negatively impact smaller models on basic reactive tasks, while larger models show more robust behaviour. Third, advanced prompting techniques primarily benefit smaller models on complex games, but offer less improvement for already high-performing large language models. Yet, we find that advanced reasoning methods yield highly variable outcomes: while capable of significantly improving performance when reasoning and decision-making align, they also introduce instability and can lead to big performance drops. Compared to human performance, our findings reveal little evidence of true emergent reasoning. Instead, large language model performance exhibits persistent limitations in areas like planning and spatial coordination, suggesting that large language models still suffer fundamental shortcomings that may not be fully overcome through self-reflective prompting alone. Reasoning is a multi-faceted task, and while methods like Chain-of-thought improve multi-step reasoning on math word problems, our findings using dynamic benchmarks highlight important shortcomings in general reasoning capabilities, indicating a need to move beyond static benchmarks to capture the complexity of reasoning.
Authors: Xuandong Zhao, Zhewei Kang, Aosong Feng, Sergey Levine, Dawn Song
Abstract: Training large language models (LLMs) for complex reasoning via Reinforcement Learning with Verifiable Rewards (RLVR) is effective but limited by reliance on costly, domain-specific supervision. We explore Reinforcement Learning from Internal Feedback (RLIF), a framework that enables LLMs to learn from intrinsic signals without external rewards or labeled data. We propose Intuitor, an RLIF method that uses a model's own confidence, termed self-certainty, as its sole reward signal. Intuitor replaces external rewards in Group Relative Policy Optimization (GRPO) with self-certainty scores, enabling fully unsupervised learning. Experiments demonstrate that Intuitor matches GRPO's performance on mathematical benchmarks while achieving superior generalization to out-of-domain tasks like code generation, without requiring gold solutions or test cases. Our findings show that intrinsic model signals can drive effective learning across domains, offering a scalable alternative to RLVR for autonomous AI systems where verifiable rewards are unavailable. Code is available at https://github.com/sunblaze-ucb/Intuitor
Authors: Tao Wu, Jingyuan Chen, Wang Lin, Mengze Li, Yumeng Zhu, Ang Li, Kun Kuang, Fei Wu
Abstract: Large language models (LLMs) are revolutionizing education, with LLM-based agents playing a key role in simulating student behavior. A major challenge in student simulation is modeling the diverse learning patterns of students at various cognitive levels. However, current LLMs, typically trained as ``helpful assistants'', target at generating perfect responses. As a result, they struggle to simulate students with diverse cognitive abilities, as they often produce overly advanced answers, missing the natural imperfections that characterize student learning and resulting in unrealistic simulations. To address this issue, we propose a training-free framework for student simulation. We begin by constructing a cognitive prototype for each student using a knowledge graph, which captures their understanding of concepts from past learning records. This prototype is then mapped to new tasks to predict student performance. Next, we simulate student solutions based on these predictions and iteratively refine them using a beam search method to better replicate realistic mistakes. To validate our approach, we construct the \texttt{Student\_100} dataset, consisting of $100$ students working on Python programming and $5,000$ learning records. Experimental results show that our method consistently outperforms baseline models, achieving $100\%$ improvement in simulation accuracy.
Authors: Guanzhou Ke, Bo Wang, Guoqing Chao, Weiming Hu, Shengfeng He
Abstract: Multimodal foundation models have demonstrated impressive capabilities across diverse tasks. However, their potential as plug-and-play solutions for missing modality reconstruction remains underexplored. To bridge this gap, we identify and formalize three potential paradigms for missing modality reconstruction, and perform a comprehensive evaluation across these paradigms, covering 42 model variants in terms of reconstruction accuracy and adaptability to downstream tasks. Our analysis reveals that current foundation models often fall short in two critical aspects: (i) fine-grained semantic extraction from the available modalities, and (ii) robust validation of generated modalities. These limitations lead to suboptimal and, at times, misaligned generations. To address these challenges, we propose an agentic framework tailored for missing modality reconstruction. This framework dynamically formulates modality-aware mining strategies based on the input context, facilitating the extraction of richer and more discriminative semantic features. In addition, we introduce a self-refinement mechanism, which iteratively verifies and enhances the quality of generated modalities through internal feedback. Experimental results show that our method reduces FID for missing image reconstruction by at least 14\% and MER for missing text reconstruction by at least 10\% compared to baselines. Code are released at: https://github.com/Guanzhou-Ke/AFM2.
Authors: Payel Bhattacharjee, Fengwei Tian, Geoffrey D. Rubin, Joseph Y. Lo, Nirav Merchant, Heidi Hanson, John Gounley, Ravi Tandon
Abstract: Purpose: This study proposes a framework for fine-tuning large language models (LLMs) with differential privacy (DP) to perform multi-abnormality classification on radiology report text. By injecting calibrated noise during fine-tuning, the framework seeks to mitigate the privacy risks associated with sensitive patient data and protect against data leakage while maintaining classification performance. Materials and Methods: We used 50,232 radiology reports from the publicly available MIMIC-CXR chest radiography and CT-RATE computed tomography datasets, collected between 2011 and 2019. Fine-tuning of LLMs was conducted to classify 14 labels from MIMIC-CXR dataset, and 18 labels from CT-RATE dataset using Differentially Private Low-Rank Adaptation (DP-LoRA) in high and moderate privacy regimes (across a range of privacy budgets = {0.01, 0.1, 1.0, 10.0}). Model performance was evaluated using weighted F1 score across three model architectures: BERT-medium, BERT-small, and ALBERT-base. Statistical analyses compared model performance across different privacy levels to quantify the privacy-utility trade-off. Results: We observe a clear privacy-utility trade-off through our experiments on 2 different datasets and 3 different models. Under moderate privacy guarantees the DP fine-tuned models achieved comparable weighted F1 scores of 0.88 on MIMIC-CXR and 0.59 on CT-RATE, compared to non-private LoRA baselines of 0.90 and 0.78, respectively. Conclusion: Differentially private fine-tuning using LoRA enables effective and privacy-preserving multi-abnormality classification from radiology reports, addressing a key challenge in fine-tuning LLMs on sensitive medical data.
Authors: Atharva Mehta, Shivam Chauhan, Monojit Choudhury
Abstract: Fine-tuning large-scale music generation models, such as MusicGen and Mustango, is a computationally expensive process, often requiring updates to billions of parameters and, therefore, significant hardware resources. Parameter-Efficient Fine-Tuning (PEFT) techniques, particularly adapter-based methods, have emerged as a promising alternative, enabling adaptation with minimal trainable parameters while preserving model performance. However, the design choices for adapters, including their architecture, placement, and size, are numerous, and it is unclear which of these combinations would produce optimal adapters and why, for a given case of low-resource music genre. In this paper, we attempt to answer this question by studying various adapter configurations for two AI music models, MusicGen and Mustango, on two genres: Hindustani Classical and Turkish Makam music. Our findings reveal distinct trade-offs: convolution-based adapters excel in capturing fine-grained local musical details such as ornamentations and short melodic phrases, while transformer-based adapters better preserve long-range dependencies crucial for structured improvisation. Additionally, we analyze computational resource requirements across different adapter scales, demonstrating how mid-sized adapters (40M parameters) achieve an optimal balance between expressivity and quality. Furthermore, we find that Mustango, a diffusion-based model, generates more diverse outputs with better adherence to the description in the input prompt while lacking in providing stability in notes, rhythm alignment, and aesthetics. Also, it is computationally intensive and requires significantly more time to train. In contrast, autoregressive models like MusicGen offer faster training and are more efficient, and can produce better quality output in comparison, but have slightly higher redundancy in their generations.
Authors: Mengyi Shan, Brian Curless, Ira Kemelmacher-Shlizerman, Steve Seitz
Abstract: We challenge text-to-image models with generating escape room puzzle images that are visually appealing, logically solid, and intellectually stimulating. While base image models struggle with spatial relationships and affordance reasoning, we propose a hierarchical multi-agent framework that decomposes this task into structured stages: functional design, symbolic scene graph reasoning, layout synthesis, and local image editing. Specialized agents collaborate through iterative feedback to ensure the scene is visually coherent and functionally solvable. Experiments show that agent collaboration improves output quality in terms of solvability, shortcut avoidance, and affordance clarity, while maintaining visual quality.
Authors: Reza Yousefi Maragheh, Pratheek Vadla, Priyank Gupta, Kai Zhao, Aysenur Inan, Kehui Yao, Jianpeng Xu, Praveen Kanumala, Jason Cho, Sushant Kumar
Abstract: Retrieval-Augmented Generation (RAG) has shown promise in enhancing recommendation systems by incorporating external context into large language model prompts. However, existing RAG-based approaches often rely on static retrieval heuristics and fail to capture nuanced user preferences in dynamic recommendation scenarios. In this work, we introduce ARAG, an Agentic Retrieval-Augmented Generation framework for Personalized Recommendation, which integrates a multi-agent collaboration mechanism into the RAG pipeline. To better understand the long-term and session behavior of the user, ARAG leverages four specialized LLM-based agents: a User Understanding Agent that summarizes user preferences from long-term and session contexts, a Natural Language Inference (NLI) Agent that evaluates semantic alignment between candidate items retrieved by RAG and inferred intent, a context summary agent that summarizes the findings of NLI agent, and an Item Ranker Agent that generates a ranked list of recommendations based on contextual fit. We evaluate ARAG accross three datasets. Experimental results demonstrate that ARAG significantly outperforms standard RAG and recency-based baselines, achieving up to 42.1% improvement in NDCG@5 and 35.5% in Hit@5. We also, conduct an ablation study to analyse the effect by different components of ARAG. Our findings highlight the effectiveness of integrating agentic reasoning into retrieval-augmented recommendation and provide new directions for LLM-based personalization.
Authors: Youkang Wang, Jian Wang, Rubing Chen, Xiao-Yong Wei
Abstract: Inference-time scaling has emerged as a powerful technique for enhancing the reasoning performance of Large Language Models (LLMs). However, existing approaches often rely on heuristic strategies for parallel sampling, lacking a principled foundation. To address this gap, we propose a probabilistic framework that formalizes the optimality of inference-time scaling under the assumption that parallel samples are independently and identically distributed (i.i.d.), and where the Best-of-N selection strategy follows a probability distribution that can be estimated. Within this framework, we derive a theoretical lower bound on the required number of samples to achieve a target performance level, providing the first principled guidance for compute-efficient scaling. Leveraging this insight, we develop OptScale, a practical algorithm that dynamically determines the optimal number of sampled responses. OptScale employs a language model-based predictor to estimate probabilistic prior parameters, enabling the decision of the minimal number of samples needed that satisfy predefined performance thresholds and confidence levels. Extensive experiments on mathematical reasoning benchmarks (including MATH-500, GSM8K, AIME, and AMC) demonstrate that OptScale significantly reduces sampling overhead while remaining better or on par with state-of-the-art reasoning performance. Our work offers both a theoretical foundation and a practical solution for principled inference-time scaling, addressing a critical gap in the efficient deployment of LLMs for complex reasoning. The source code is publicly available at https://github.com/Albertwyk/OptScale.
Authors: Xiao Zhang, Johan Bos
Abstract: Tombstones are historically and culturally rich artifacts, encapsulating individual lives, community memory, historical narratives and artistic expression. Yet, many tombstones today face significant preservation challenges, including physical erosion, vandalism, environmental degradation, and political shifts. In this paper, we introduce a novel multi-modal framework for tombstones digitization, aiming to improve the interpretation, organization and retrieval of tombstone content. Our approach leverages vision-language models (VLMs) to translate tombstone images into structured Tombstone Meaning Representations (TMRs), capturing both image and text information. To further enrich semantic parsing, we incorporate retrieval-augmented generation (RAG) for integrate externally dependent elements such as toponyms, occupation codes, and ontological concepts. Compared to traditional OCR-based pipelines, our method improves parsing accuracy from an F1 score of 36.1 to 89.5. We additionally evaluate the model's robustness across diverse linguistic and cultural inscriptions, and simulate physical degradation through image fusion to assess performance under noisy or damaged conditions. Our work represents the first attempt to formalize tombstone understanding using large vision-language models, presenting implications for heritage preservation.
Authors: Shuquan Lian, Yuhang Wu, Jia Ma, Yifan Ding, Zihan Song, Bingqi Chen, Xiawu Zheng, Hui Li
Abstract: The emergence of Multimodal Large Language Models (MLLMs) has driven significant advances in Graphical User Interface (GUI) agent capabilities. Nevertheless, existing GUI agent training and inference techniques still suffer from a dilemma for reasoning designs, ineffective reward, and visual noise. To address these issues, we introduce UI-AGILE for enhancing GUI agents at both training and inference. For training, we propose a suite of improvements to the Supervised Fine-Tuning (SFT) process: 1) a continuous reward function to incentivize high-precision grounding; 2) a ``Simple Thinking'' reward to balance planning with speed and grounding accuracy; and 3) a cropping-based resampling strategy to mitigate the sparse reward problem and improve learning on complex tasks. For inference, we present decomposed grounding with selection to dramatically improve grounding accuracy on high-resolution displays by breaking the image into smaller, manageable parts. Experiments show that UI-AGILE achieves the state-of-the-art grounding performance on two benchmarks ScreenSpot-Pro and ScreenSpot-v2 while it also exhibits strong general agent capabilities. For instance, using both our training and inference enhancement methods brings 23\% grounding accuracy improvement over the best baseline on ScreenSpot-Pro. We provide the code in https://github.com/KDEGroup/UI-AGILE.
Authors: Long Phan, Mantas Mazeika, Andy Zou, Dan Hendrycks
Abstract: Evaluating AI agents within complex, interactive environments that mirror real-world challenges is critical for understanding their practical capabilities. While existing agent benchmarks effectively assess skills like tool use or performance on structured tasks, they often do not fully capture an agent's ability to operate autonomously in exploratory environments that demand sustained, self-directed reasoning over a long and growing context. To spur the development of agents capable of more robust intrinsic reasoning over long horizons, we introduce TextQuests, a benchmark based on the Infocom suite of interactive fiction games. These text-based adventures, which can take human players over 30 hours and require hundreds of precise actions to solve, serve as an effective proxy for evaluating AI agents on focused, stateful tasks. The benchmark is specifically designed to assess an LLM agent's capacity for self-contained problem-solving by precluding the use of external tools, thereby focusing on intrinsic long-context reasoning capabilities in an exploratory environment characterized by the need for trial-and-error learning and sustained problem-solving within a single interactive session. We release TextQuests at https://textquests.ai.
URLs: https://textquests.ai.