Authors: Jiahong Li, Yiwen Shao, Jianheng Zhuo, Chenda Li, Liliang Tang, Dong Yu, Yanmin Qian
Abstract: Recent advancements in deep learning have significantly enhanced multilingual automatic speech recognition (ASR) due to the development of advanced model architectures and available large-scale multilingual datasets. Despite that, multilingual ASR still suffers from the curse of multilinguality in that different languages tend to interfere with each other, making it difficult for the ASR model to identify multiple languages effectively while sharing model capacity across them. This paper proposes an efficient finetuning framework for customized multilingual ASR via prepared LoRA language experts based on Whisper. Through LoRA expert fusion or knowledge distillation, our approach achieves better recognition performance on target languages than standard fine-tuning methods. Experimental results demonstrate that the proposed models yield approximately 10\% and 15\% relative performance gains in language-aware and language-agnostic scenarios, respectively.
Authors: Hyeongcheol Park, MinHyuk Jang, Ha Dam Baek, Gyusam Chang, Jiyoung Seo, Jiwan Park, Hogun Park, Sangpil Kim
Abstract: Multimodal Knowledge Graphs (MMKGs), which represent explicit knowledge across multiple modalities, play a pivotal role by complementing the implicit knowledge of Multimodal Large Language Models (MLLMs) and enabling more grounded reasoning via Retrieval Augmented Generation (RAG). However, existing MMKGs are generally limited in scope: they are often constructed by augmenting pre-existing knowledge graphs, which restricts their knowledge, resulting in outdated or incomplete knowledge coverage, and they often support only a narrow range of modalities, such as text and visual information. These limitations reduce their extensibility and applicability to a broad range of multimodal tasks, particularly as the field shifts toward richer modalities such as video and audio in recent MLLMs. Therefore, we propose the Visual-Audio-Text Knowledge Graph (VAT-KG), the first concept-centric and knowledge-intensive multimodal knowledge graph that covers visual, audio, and text information, where each triplet is linked to multimodal data and enriched with detailed descriptions of concepts. Specifically, our construction pipeline ensures cross-modal knowledge alignment between multimodal data and fine-grained semantics through a series of stringent filtering and alignment steps, enabling the automatic generation of MMKGs from any multimodal dataset. We further introduce a novel multimodal RAG framework that retrieves detailed concept-level knowledge in response to queries from arbitrary modalities. Experiments on question answering tasks across various modalities demonstrate the effectiveness of VAT-KG in supporting MLLMs, highlighting its practical value in unifying and leveraging multimodal knowledge.
Authors: Kaiying Yan, Moyang Liu, Yukun Liu, Ruibo Fu, Zhengqi Wen, Jianhua Tao, Xuefei Liu
Abstract: The rapid spread of fake news across multimedia platforms presents serious challenges to information credibility. In this paper, we propose a Debunk-and-Infer framework for Fake News Detection(DIFND) that leverages debunking knowledge to enhance both the performance and interpretability of fake news detection. DIFND integrates the generative strength of conditional diffusion models with the collaborative reasoning capabilities of multimodal large language models (MLLMs). Specifically, debunk diffusion is employed to generate refuting or authenticating evidence based on the multimodal content of news videos, enriching the evaluation process with diverse yet semantically aligned synthetic samples. To improve inference, we propose a chain-of-debunk strategy where a multi-agent MLLM system produces logic-grounded, multimodal-aware reasoning content and final veracity judgment. By jointly modeling multimodal features, generative debunking cues, and reasoning-rich verification within a unified architecture, DIFND achieves notable improvements in detection accuracy. Extensive experiments on the FakeSV and FVC datasets show that DIFND not only outperforms existing approaches but also delivers trustworthy decisions.
Authors: FutureSearch, :, Jack Wildman, Nikos I. Bosse, Daniel Hnyk, Peter M\"uhlbacher, Finn Hambly, Jon Evans, Dan Schwarz, Lawrence Phillips
Abstract: Forecasting is a challenging task that offers a clearly measurable way to study AI systems. Forecasting requires a large amount of research on the internet, and evaluations require time for events to happen, making the development of forecasting benchmarks challenging. To date, no forecasting benchmark provides a realistic, hermetic, and repeatable environment for LLM forecasters. We introduce Bench To the Future (BTF), a "pastcasting" benchmark with hundreds of high-quality questions for which the resolution is already known. Each question is accompanied by a large offline corpus of tens of thousands of relevant web pages, enabling a way to elicit realistic "forecasts" on past events from LLMs. Results suggest that our pastcasting environment can produce results comparable to those based on forecasts using the internet on at-the-time unresolved questions. We show results benchmarking agent and chain-of-thought forecasting approaches using several LLMs, including the recently-released Claude 4 models, and demonstrate BTF's ability to track steady forecasting capability progress over time. We intend this to be a living benchmark, with new questions added continually to account for increasing training data cutoff dates. We invite researchers to contact us at hello@futuresearch.ai to utilize our benchmark or tooling for their own research.
Authors: Junze Chen, Cheng Yang, Shujie Li, Zhiqiang Zhang, Yawen Li, Junping Du, Chuan Shi
Abstract: Large language models (LLMs) have demonstrated their strong capabilities in various domains, and have been recently integrated for graph analysis as graph language models (GLMs). With LLMs as the predictor, some GLMs can interpret unseen tasks described by natural language, and learn from a few examples in the prompts without parameter tuning, known as in-context learning (ICL). Another subset of GLMs utilizes abundant training labels to enhance model performance, known as instruction tuning. However, we argue that ICL on graphs has effectiveness issues due to fixed parameters and efficiency issues due to long context. Meanwhile, the large amount of labeled data required for instruction tuning can be difficult to obtain in real-world scenarios. To this end, we aim to introduce an extra parameter adaptation stage that can efficiently tailor GLMs to an unseen graph and task with only a few labeled examples, in exchange for better prediction accuracy and faster inference speed. For implementation, in this paper we propose GraphLAMA method, with its model backbone and learning schemes specialized for efficient tuning and inference. Specifically, for model backbone, we use a graph neural network (GNN) with several well-designed components to transform nodes into the representation space of LLM tokens. Task instructions can then be represented as a mixture of node and language tokens. In the pre-training stage, model parameters except the LLM will be trained with different tasks to capture general knowledge. In the adaptation stage, only a few pre-trained parameters will be updated based on few-shot examples. Extensive experiments on few/zero-shot node classification and summary generation show that our proposed GraphLAMA achieves state-of-the-art performance with 4.91% absolution improvement in accuracy. Compared with ICL, our inference speed can be 10 times faster under 5-shot setting.
Authors: Yifu Han, Geo Zhang
Abstract: This study investigates the effectiveness of reinforcement learning (RL) fine-tuning techniques on a compact language model (Qwen2.5-0.5B Base) for two challenging tasks: instruction following and mathematical reasoning. We compare supervised fine-tuning (SFT), Direct Preference Optimization (DPO) using preference-labeled data, and Reinforce Leave-One-Out (RLOO) with reward models. Our experiments show that RLOO with DeBERTa reward modeling achieves the best alignment, while DPO provides strong and consistent results. For math reasoing tasks, synthetic data augmentation and best-of-N sampling with an external verifier significantly improve accuracy, showing the potential of combining fine-tuning with inference-time tools. This study highlights key trade-offs and practical strategies for training lightweight, task-aligned small-scale language models.
Authors: Emilio Barkett, Olivia Long, Madhavendra Thakur
Abstract: Despite their widespread use in fact-checking, moderation, and high-stakes decision-making, large language models (LLMs) remain poorly understood as judges of truth. This study presents the largest evaluation to date of LLMs' veracity detection capabilities and the first analysis of these capabilities in reasoning models. We had eight LLMs make 4,800 veracity judgments across several prompts, comparing reasoning and non-reasoning models. We find that rates of truth-bias, or the likelihood to believe a statement is true, regardless of whether it is actually true, are lower in reasoning models than in non-reasoning models, but still higher than human benchmarks. Most concerning, we identify sycophantic tendencies in several advanced models (o4-mini and GPT-4.1 from OpenAI, R1 from DeepSeek), which displayed an asymmetry in detection accuracy, performing well in truth accuracy but poorly in deception accuracy. This suggests that capability advances alone do not resolve fundamental veracity detection challenges in LLMs.
Authors: Jun Yin, Pengyu Zeng, Jing Zhong, Peilin Li, Miao Zhang, Ran Luo, Shuai Lu
Abstract: In the architectural design process, floor plan generation is inherently progressive and iterative. However, existing generative models for floor plans are predominantly end-to-end generation that produce an entire pixel-based layout in a single pass. This paradigm is often incompatible with the incremental workflows observed in real-world architectural practice. To address this issue, we draw inspiration from the autoregressive 'next token prediction' mechanism commonly used in large language models, and propose a novel 'next room prediction' paradigm tailored to architectural floor plan modeling. Experimental evaluation indicates that FPDS demonstrates competitive performance in comparison to diffusion models and Tell2Design in the text-to-floorplan task, indicating its potential applicability in supporting future intelligent architectural design.
Authors: Kaiying Kevin Lin, Hsiyu Chen, Haopeng Zhang
Abstract: While large language models (LLMs) have demonstrated impressive performance across a wide range of natural language processing (NLP) tasks in high-resource languages, their capabilities in low-resource and minority languages remain significantly underexplored. Formosan languages -- a subgroup of Austronesian languages spoken in Taiwan -- are both linguistically rich and endangered, largely due to the sociolinguistic dominance of Mandarin. In this work, we introduce FORMOSANBENCH, the first benchmark for evaluating LLMs on low-resource Austronesian languages. It covers three endangered Formosan languages: Atayal, Amis, and Paiwan, across three core NLP tasks: machine translation, automatic speech recognition (ASR), and text summarization. We assess model performance in zero-shot, 10-shot, and fine-tuned settings using FORMOSANBENCH. Our results reveal a substantial performance gap between high-resource and Formosan languages. Existing LLMs consistently underperform across all tasks, with 10-shot learning and fine-tuning offering only limited improvements. These findings underscore the urgent need for more inclusive NLP technologies that can effectively support endangered and underrepresented languages. We release our datasets and code to facilitate future research in this direction.
Authors: Jiyan Liu, Youzheng Liu, Taihang Wang, Xiaoman Xu, Yimin Wang, Ye Jiang
Abstract: This paper describes the participation of QUST_NLP in the SemEval-2025 Task 7. We propose a three-stage retrieval framework specifically designed for fact-checked claim retrieval. Initially, we evaluate the performance of several retrieval models and select the one that yields the best results for candidate retrieval. Next, we employ multiple re-ranking models to enhance the candidate results, with each model selecting the Top-10 outcomes. In the final stage, we utilize weighted voting to determine the final retrieval outcomes. Our approach achieved 5th place in the monolingual track and 7th place in the crosslingual track. We release our system code at: https://github.com/warmth27/SemEval2025_Task7.
Authors: Takato Ueno, Keito Inoshita
Abstract: Japan's kairanban culture and idobata conversations have long functioned as traditional communication practices that foster nuanced dialogue among community members and contribute to the formation of social balance. Inspired by these information exchange processes, this study proposes a multi-agent inference framework (KCS+IBC) that integrates multiple large language models (LLMs) to achieve bias mitigation, improved explainability, and probabilistic prediction in sentiment analysis. In addition to sequentially sharing prediction results, the proposed method incorporates a mid-phase casual dialogue session to blend formal inference with individual perspectives and introduces probabilistic sentiment prediction. Experimental results show that KCS achieves accuracy comparable to that of a single LLM across datasets, while KCS+IBC exhibits a consistent decrease in entropy and a gradual increase in variance during the latter stages of inference, suggesting the framework's ability to balance aggregation and diversity of predictions. Future work will quantitatively assess the impact of these characteristics on bias correction and aim to develop more advanced sentiment analysis systems.
Authors: Arwa Arif
Abstract: Backtranslation BT is widely used in low resource machine translation MT to generate additional synthetic training data using monolingual corpora. While this approach has shown strong improvements for many language pairs, its effectiveness in high quality, low resource settings remains unclear. In this work, we explore the effectiveness of backtranslation for English Gujarati translation using the multilingual pretrained MBART50 model. Our baseline system, trained on a high quality parallel corpus of approximately 50,000 sentence pairs, achieves a BLEU score of 43.8 on a validation set. We augment this data with carefully filtered backtranslated examples generated from monolingual Gujarati text. Surprisingly, adding this synthetic data does not improve translation performance and, in some cases, slightly reduces it. We evaluate our models using multiple metrics like BLEU, ChrF++, TER, BLEURT and analyze possible reasons for this saturation. Our findings suggest that backtranslation may reach a point of diminishing returns in certain low-resource settings and we discuss implications for future research.
Authors: Baqer M. Merzah, Tania Taami, Salman Asoudeh, Amir reza Hossein pour, Saeed Mirzaee, Amir Ali Bengari
Abstract: Large Language Models (LLMs) have recently gained attention in the life sciences due to their capacity to model, extract, and apply complex biological information. Beyond their classical use as chatbots, these systems are increasingly used for complex analysis and problem-solving in specialized fields, including bioinformatics. First, we introduce BIOPARS-BENCH, a dataset from over 10,000 scientific articles, textbooks, and medical websites. BioParsQA was also introduced to evaluate the proposed model, which consists of 5,231 Persian medical questions and answers. This study then introduces BioPars, a simple but accurate measure designed to assess LLMs for three main abilities: acquiring subject-specific knowledge, interpreting and synthesizing such knowledge, and demonstrating proper evidence. Comparing ChatGPT, Llama, and Galactica, our study highlights their ability to remember and retrieve learned knowledge but also reveals shortcomings in addressing higher-level, real-world questions and fine-grained inferences. These findings indicate the need for further fine-tuning to address the capabilities of LLM in bioinformatics tasks. To our knowledge, BioPars is the first application of LLM in Persian medical QA, especially for generating long answers. Evaluation of four selected medical QA datasets shows that BioPars has achieved remarkable results compared to comparative approaches. The model on BioParsQA achieved a ROUGE-L score of 29.99, which is an improvement over GPT-4 1.0. The model achieved a BERTScore of 90.87 with the MMR method. The MoverScore and BLEURT values were also higher in this model than the other three models. In addition, the reported scores for the model are MoverScore=60.43 and BLEURT=50.78. BioPars is an ongoing project and all resources related to its development will be made available via the following GitHub repository: https://github.com/amirap80/BioPars.
Authors: Andrejs Sorstkins
Abstract: Resource efficiency is a critical barrier to deploying large language models (LLMs) in edge and privacy-sensitive applications. This study evaluates the efficacy of two augmentation strategies--Retrieval-Augmented Generation (RAG) and Hypothetical Document Embeddings (HyDE)--on compact Gemma LLMs of 1 billion and 4 billion parameters, within the context of a privacy-first personal assistant. We implement short-term memory via MongoDB and long-term semantic storage via Qdrant, orchestrated through FastAPI and LangChain, and expose the system through a React.js frontend. Across both model scales, RAG consistently reduces latency by up to 17\% and eliminates factual hallucinations when responding to user-specific and domain-specific queries. HyDE, by contrast, enhances semantic relevance--particularly for complex physics prompts--but incurs a 25--40\% increase in response time and a non-negligible hallucination rate in personal-data retrieval. Comparing 1 B to 4 B models, we observe that scaling yields marginal throughput gains for baseline and RAG pipelines, but magnifies HyDE's computational overhead and variability. Our findings position RAG as the pragmatic choice for on-device personal assistants powered by small-scale LLMs.
Authors: Weihua Xiao, Derek Ekberg, Siddharth Garg, Ramesh Karri
Abstract: SystemVerilog Assertions (SVAs) are critical for verifying the correctness of hardware designs, but manually writing them from natural language property descriptions, i.e., NL2SVA, remains a labor-intensive and error-prone task. Recent advances in large language models (LLMs) offer opportunities to automate this translation. However, existing models still struggle with understanding domain-specific syntax and semantics. To enhance LLM performance in NL2SVA, we propose a customized retrieval-augmented generation (RAG) framework and a synthetic fine-tuning dataset that together improve LLM's performance. To further improve lightweight models over NL2SVA, our fine-tuning dataset provides prompt-guided explanations that teach LLMs the layer-by-layer construction process of concurrent SVAs, enabling supervised fine-tuning that greatly improves syntax and functionality accuracy. To evaluate the performance of LLMs over NL2SVA, we construct the largest evaluation dataset for NL2SVA, comprising 40 Verilog designs and 229 formally verified SVAs with detailed annotations. Experimental results show that our customized RAG framework increases the number of functionality matched SVAs by 58.42% over GPT-4o-mini, while Qwen2.5-Coder-7B-Instruct fine-tuned on our fine-tuning dataset and integrated with HybridRetrieval achieves a 59.05% over the base Qwen model.
Authors: Roland Riachi, Kashif Rasul, Arjun Ashok, Prateek Humane, Alexis Roger, Andrew R. Williams, Yuriy Nevmyvaka, Irina Rish
Abstract: Recent works have demonstrated the effectiveness of adapting pre-trained language models (LMs) for forecasting time series in the low-data regime. We build upon these findings by analyzing the effective transfer from language models to time series forecasting under various design choices including upstream post-training, time series tokenizer and language backbone size. In the low-data regime, these design choices have a significant impact on the validation loss, with clear-cut choices that outperform others. Contrary to Hernandez et al. (2021), we observe that the validation loss of the LMs continues to smoothly decrease long after the validation loss of the randomly initialized models has converged, leading to a non-vanishing transfer gap that holds across design choices. These findings not only help shed light on the effective use of compute-efficient training for time series, but also open the way for the study of modality-agnostic properties of data distributions leveraged by these models.
Authors: Jianshuo Dong, Yujia Fu, Chuanrui Hu, Chao Zhang, Han Qiu
Abstract: Large Reasoning Models (LRMs), which autonomously produce a reasoning Chain of Thought (CoT) before producing final responses, offer a promising approach to interpreting and monitoring model behaviors. Inspired by the observation that certain CoT patterns -- e.g., ``Wait, did I miss anything?'' -- consistently emerge across tasks, we explore whether LRMs exhibit human-like cognitive habits. Building on Habits of Mind, a well-established framework of cognitive habits associated with successful human problem-solving, we introduce CogTest, a principled benchmark designed to evaluate LRMs' cognitive habits. CogTest includes 16 cognitive habits, each instantiated with 25 diverse tasks, and employs an evidence-first extraction method to ensure reliable habit identification. With CogTest, we conduct a comprehensive evaluation of 16 widely used LLMs (13 LRMs and 3 non-reasoning ones). Our findings reveal that LRMs, unlike conventional LLMs, not only exhibit human-like habits but also adaptively deploy them according to different tasks. Finer-grained analyses further uncover patterns of similarity and difference in LRMs' cognitive habit profiles, particularly certain inter-family similarity (e.g., Qwen-3 models and DeepSeek-R1). Extending the study to safety-related tasks, we observe that certain habits, such as Taking Responsible Risks, are strongly associated with the generation of harmful responses. These findings suggest that studying persistent behavioral patterns in LRMs' CoTs is a valuable step toward deeper understanding of LLM misbehavior. The code is available at: https://github.com/jianshuod/CogTest.
Authors: Tianyu. Zou, Shengwu. Xiong, Ruilin. Yao, Jirui. Huang, Yi. Rong, Yaxiong. Chen, Shili. Xiong, Cong. Wang
Abstract: Evaluating multimodal large language models (MLLMs) remains a fundamental challenge due to a lack of structured, interpretable, and theoretically grounded benchmark designs. Existing benchmarks often adopt heuristic-based task groupings with unclear cognitive targets, thus resulting in overlapping abilities, redundant indicators, and limited diagnostic power. In this work, we propose a novel framework for aligning MLLM benchmark based on Structural Equation Modeling (SEM) to analyze and quantify the internal validity, dimensional separability, and contribution of benchmark components. Motivated by the observed limitations of current designs, we further introduce a novel capability hierarchy grounded in Piagets theory of cognitive development, dividing MLLM abilities into three hierarchical layers, i.e., Perception, Memory, and Reasoning. We reorganize existing MLLM benchmarks under the proposed framework and construct a new benchmark named Gold. Experimental results demonstrate that the proposed benchmark exhibits stronger interpretability, reduced indicator redundancy, and clearer cognitive consistency compared to existing approaches.
Authors: Yanwei Ren, Liu Liu, Baosheng Yu, Jiayan Qiu, Quan Chen
Abstract: Optimizing instructions for large language models (LLMs) is critical for harnessing their full potential in complex and diverse tasks. However, relying solely on white-box approaches demands extensive computational resources and offers limited representational capacity, while black-box models can incur prohibitive financial costs. To address these challenges, we introduce a novel framework that seamlessly merges the strengths of both paradigms. Black-box models provide high-quality, diverse instruction initializations, and white-box models supply fine-grained interpretability through hidden states and output features. By enforcing a semantic similarity constraint, these components fuse into a unified high-dimensional representation that captures deep semantic and structural nuances, enabling an iterative optimization process to refine instruction quality and adaptability. Extensive evaluations across a broad spectrum of tasks-ranging from complex reasoning to cross-lingual generalization-demonstrate that our approach consistently outperforms state-of-the-art baselines. This fusion of black-box initialization with advanced semantic refinement yields a scalable and efficient solution, paving the way for next-generation LLM-driven applications in diverse real-world scenarios. The source code will be released soon.
Authors: Yicheng Mao, Yang Zhao
Abstract: With globalization and increasing immigrant populations, immigration departments face significant work-loads and the challenge of ensuring fairness in decision-making processes. Integrating artificial intelligence offers a promising solution to these challenges. This study investigates the potential of large language models (LLMs),such as GPT-3.5 and GPT-4, in supporting immigration decision-making. Utilizing a mixed-methods approach,this paper conducted discrete choice experiments and in-depth interviews to study LLM decision-making strategies and whether they are fair. Our findings demonstrate that LLMs can align their decision-making with human strategies, emphasizing utility maximization and procedural fairness. Meanwhile, this paper also reveals that while ChatGPT has safeguards to prevent unintentional discrimination, it still exhibits stereotypes and biases concerning nationality and shows preferences toward privileged group. This dual analysis highlights both the potential and limitations of LLMs in automating and enhancing immigration decisions.
Authors: Josefa Lia Stoisser, Marc Boubnovski Martell, Lawrence Phillips, Casper Hansen, Julien Fauqueur
Abstract: We propose STRuCT-LLM, a unified framework for training large language models (LLMs) to perform structured reasoning over both relational and graph-structured data. Our approach jointly optimizes Text-to-SQL and Text-to-Cypher tasks using reinforcement learning (RL) combined with Chain-of-Thought (CoT) supervision. To support fine-grained optimization in graph-based parsing, we introduce a topology-aware reward function based on graph edit distance. Unlike prior work that treats relational and graph formalisms in isolation, STRuCT-LLM leverages shared abstractions between SQL and Cypher to induce cross-formalism transfer, enabling SQL training to improve Cypher performance and vice versa - even without shared schemas. Our largest model (QwQ-32B) achieves substantial relative improvements across tasks: on semantic parsing, Spider improves by 13.5\% and Text2Cypher by 73.1\%. The model also demonstrates strong zero-shot generalization, improving performance on downstream tabular QA (TableBench: 8.5\%) and knowledge graph QA (CR-LT-KGQA: 1.7\%) without any QA-specific supervision. These results demonstrate both the effectiveness of executable queries as scaffolds for structured reasoning and the synergistic benefits of jointly training on SQL and Cypher (code available at https://github.com/bouv/STRuCT-LLM).
Authors: Hongli Yang, Yizhou Peng, Hao Huang, Sheng Li
Abstract: Large-scale multilingual ASR models like Whisper excel in high-resource settings but face challenges in low-resource scenarios, such as rare languages and code-switching (CS), due to computational costs and catastrophic forgetting. We explore Soft Prompt Tuning (SPT), a parameter-efficient method to enhance CS ASR while preserving prior knowledge. We evaluate two strategies: (1) full fine-tuning (FFT) of both soft prompts and the entire Whisper model, demonstrating improved cross-lingual capabilities compared to traditional methods, and (2) adhering to SPT's original design by freezing model parameters and only training soft prompts. Additionally, we introduce SPT4ASR, a combination of different SPT variants. Experiments on the SEAME and ASRU2019 datasets show that deep prompt tuning is the most effective SPT approach, and our SPT4ASR methods achieve further error reductions in CS ASR, maintaining parameter efficiency similar to LoRA, without degrading performance on existing languages.
Authors: Hongli Yang, Sheng Li, Hao Huang, Ayiduosi Tuohan, Yizhou Peng
Abstract: Recent advancements in multilingual automatic speech recognition (ASR) have been driven by large-scale end-to-end models like Whisper. However, challenges such as language interference and expanding to unseen languages (language expansion) without degrading performance persist. This paper addresses these with three contributions: 1) Entire Soft Prompt Tuning (Entire SPT), which applies soft prompts to both the encoder and decoder, enhancing feature extraction and decoding; 2) Language-Aware Prompt Tuning (LAPT), which leverages cross-lingual similarities to encode shared and language-specific features using lightweight prompt matrices; 3) SPT-Whisper, a toolkit that integrates SPT into Whisper and enables efficient continual learning. Experiments across three languages from FLEURS demonstrate that Entire SPT and LAPT outperform Decoder SPT by 5.0% and 16.0% in language expansion tasks, respectively, providing an efficient solution for dynamic, multilingual ASR models with minimal computational overhead.
Authors: Andrew Maranh\~ao Ventura D'addario
Abstract: The evaluation of Large Language Models (LLMs) in healthcare has been dominated by physician-centric, English-language benchmarks, creating a dangerous illusion of competence that ignores the interprofessional nature of patient care. To provide a more holistic and realistic assessment, we introduce HealthQA-BR, the first large-scale, system-wide benchmark for Portuguese-speaking healthcare. Comprising 5,632 questions from Brazil's national licensing and residency exams, it uniquely assesses knowledge not only in medicine and its specialties but also in nursing, dentistry, psychology, social work, and other allied health professions. We conducted a rigorous zero-shot evaluation of over 20 leading LLMs. Our results reveal that while state-of-the-art models like GPT 4.1 achieve high overall accuracy (86.6%), this top-line score masks alarming, previously unmeasured deficiencies. A granular analysis shows performance plummets from near-perfect in specialties like Ophthalmology (98.7%) to barely passing in Neurosurgery (60.0%) and, most notably, Social Work (68.4%). This "spiky" knowledge profile is a systemic issue observed across all models, demonstrating that high-level scores are insufficient for safety validation. By publicly releasing HealthQA-BR and our evaluation suite, we provide a crucial tool to move beyond single-score evaluations and toward a more honest, granular audit of AI readiness for the entire healthcare team.
Authors: Dana Alsagheer, Yang Lu, Abdulrahman Kamal, Omar Kamal, Mohammad Kamal, Nada Mansour, Cosmo Yang Wu, Rambiba Karanjai, Sen Li, Weidong Shi
Abstract: Recent advancements in Large Language Models (LLMs) have demonstrated remarkable capabilities in various domains. However, effective decision-making relies heavily on strong reasoning abilities. Reasoning is the foundation for decision-making, providing the analytical and logical framework to make sound choices. Reasoning involves analyzing information, drawing inferences, and reaching conclusions based on logic or evidence. Decision-making builds on this foundation by applying the insights from reasoning to select the best course of action among alternatives. Together, these processes create a continuous cycle of thought and action aimed at achieving goals effectively. As AI technology evolves, there is a growing trend to train LLMs to excel in general reasoning. This study explores how the general reasoning capabilities of LLMs connect to their performance in domain-specific reasoning tasks.
Authors: Sam Yu-Te Lee, Chengyang Ji, Shicheng Wen, Lifu Huang, Dongyi Liu, Kwan-Liu Ma
Abstract: Text analytics has traditionally required specialized knowledge in Natural Language Processing (NLP) or text analysis, which presents a barrier for entry-level analysts. Recent advances in large language models (LLMs) have changed the landscape of NLP by enabling more accessible and automated text analysis (e.g., topic detection, summarization, information extraction, etc.). We introduce VIDEE, a system that supports entry-level data analysts to conduct advanced text analytics with intelligent agents. VIDEE instantiates a human-agent collaroration workflow consisting of three stages: (1) Decomposition, which incorporates a human-in-the-loop Monte-Carlo Tree Search algorithm to support generative reasoning with human feedback, (2) Execution, which generates an executable text analytics pipeline, and (3) Evaluation, which integrates LLM-based evaluation and visualizations to support user validation of execution results. We conduct two quantitative experiments to evaluate VIDEE's effectiveness and analyze common agent errors. A user study involving participants with varying levels of NLP and text analytics experience -- from none to expert -- demonstrates the system's usability and reveals distinct user behavior patterns. The findings identify design implications for human-agent collaboration, validate the practical utility of VIDEE for non-expert users, and inform future improvements to intelligent text analytics systems.
Authors: Muhammad Ahmad, Muhammad Waqas, Ameer Hamza, Ildar Batyrshin, Grigori Sidorov
Abstract: Hope is a positive emotional state involving the expectation of favorable future outcomes, while hope speech refers to communication that promotes optimism, resilience, and support, particularly in adverse contexts. Although hope speech detection has gained attention in Natural Language Processing (NLP), existing research mainly focuses on high-resource languages and standardized scripts, often overlooking informal and underrepresented forms such as Roman Urdu. To the best of our knowledge, this is the first study to address hope speech detection in code-mixed Roman Urdu by introducing a carefully annotated dataset, thereby filling a critical gap in inclusive NLP research for low-resource, informal language varieties. This study makes four key contributions: (1) it introduces the first multi-class annotated dataset for Roman Urdu hope speech, comprising Generalized Hope, Realistic Hope, Unrealistic Hope, and Not Hope categories; (2) it explores the psychological foundations of hope and analyzes its linguistic patterns in code-mixed Roman Urdu to inform dataset development; (3) it proposes a custom attention-based transformer model optimized for the syntactic and semantic variability of Roman Urdu, evaluated using 5-fold cross-validation; and (4) it verifies the statistical significance of performance gains using a t-test. The proposed model, XLM-R, achieves the best performance with a cross-validation score of 0.78, outperforming the baseline SVM (0.75) and BiLSTM (0.76), with gains of 4% and 2.63% respectively.
Authors: J. Koorndijk
Abstract: Current literature suggests that alignment faking (deceptive alignment) is an emergent property of large language models. We present the first empirical evidence that a small instruction-tuned model, specifically LLaMA 3 8B, can also exhibit alignment faking. We further show that prompt-only interventions, including deontological moral framing and scratchpad reasoning, significantly reduce this behavior without modifying model internals. This challenges the assumption that prompt-based ethics are trivial and that deceptive alignment requires scale. We introduce a taxonomy distinguishing shallow deception, shaped by context and suppressible through prompting, from deep deception, which reflects persistent, goal-driven misalignment. Our findings refine the understanding of deception in language models and underscore the need for alignment evaluations across model sizes and deployment settings.
Authors: Christoph Brosch, Sian Brumm, Rolf Krieger, Jonas Scheffler
Abstract: Generative AI and large language models (LLMs) offer significant potential for automating the extraction of structured information from web pages. In this work, we focus on food product pages from online retailers and explore schema-constrained extraction approaches to retrieve key product attributes, such as ingredient lists and nutrition tables. We compare two LLM-based approaches, direct extraction and indirect extraction via generated functions, evaluating them in terms of accuracy, efficiency, and cost on a curated dataset of 3,000 food product pages from three different online shops. Our results show that although the indirect approach achieves slightly lower accuracy (96.48\%, $-1.61\%$ compared to direct extraction), it reduces the number of required LLM calls by 95.82\%, leading to substantial efficiency gains and lower operational costs. These findings suggest that indirect extraction approaches can provide scalable and cost-effective solutions for large-scale information extraction tasks from template-based web pages using LLMs.
Authors: Hyundong Cho, Spencer Lin, Tejas Srinivasan, Michael Saxon, Deuksin Kwon, Natali T. Chavez, Jonathan May
Abstract: Nonverbal communication (NVC) plays an integral role in human language, but studying NVC in general is challenging because of its broad scope and high variance in interpretation among individuals and cultures. However, mime -- the theatrical technique of suggesting intent using only gesture, expression, and movement -- is a subset of NVC that consists of explicit and embodied actions with much lower human interpretation variance. We argue that a solid understanding of mimed actions is a crucial prerequisite for vision-language models capable of interpreting and commanding more subtle aspects of NVC. Hence, we propose Mime Identification Multimodal Evaluation (MIME), a novel video-based question answering benchmark comprising of 86 mimed actions. Constructed with motion capture data, MIME consists of variations of each action with perturbations applied to the character, background, and viewpoint for evaluating recognition robustness. We find that both open-weight and API-based vision-language models perform significantly worse than humans on MIME, motivating the need for increased research for instilling more robust understanding of human gestures.
Authors: Weihong Qi, Fan Huang, Jisun An, Haewoon Kwak
Abstract: This study evaluates the ability of DeepSeek, an open-source large language model (LLM), to simulate public opinions in comparison to LLMs developed by major tech companies. By comparing DeepSeek-R1 and DeepSeek-V3 with Qwen2.5, GPT-4o, and Llama-3.3 and utilizing survey data from the American National Election Studies (ANES) and the Zuobiao dataset of China, we assess these models' capacity to predict public opinions on social issues in both China and the United States, highlighting their comparative capabilities between countries. Our findings indicate that DeepSeek-V3 performs best in simulating U.S. opinions on the abortion issue compared to other topics such as climate change, gun control, immigration, and services for same-sex couples, primarily because it more accurately simulates responses when provided with Democratic or liberal personas. For Chinese samples, DeepSeek-V3 performs best in simulating opinions on foreign aid and individualism but shows limitations in modeling views on capitalism, particularly failing to capture the stances of low-income and non-college-educated individuals. It does not exhibit significant differences from other models in simulating opinions on traditionalism and the free market. Further analysis reveals that all LLMs exhibit the tendency to overgeneralize a single perspective within demographic groups, often defaulting to consistent responses within groups. These findings highlight the need to mitigate cultural and demographic biases in LLM-driven public opinion modeling, calling for approaches such as more inclusive training methodologies.
Authors: Ilya Lasy, Peter Knees, Stefan Woltran
Abstract: Underlying mechanisms of memorization in LLMs -- the verbatim reproduction of training data -- remain poorly understood. What exact part of the network decides to retrieve a token that we would consider as start of memorization sequence? How exactly is the models' behaviour different when producing memorized sentence vs non-memorized? In this work we approach these questions from mechanistic interpretability standpoint by utilizing transformer circuits -- the minimal computational subgraphs that perform specific functions within the model. Through carefully constructed contrastive datasets, we identify points where model generation diverges from memorized content and isolate the specific circuits responsible for two distinct aspects of memorization. We find that circuits that initiate memorization can also maintain it once started, while circuits that only maintain memorization cannot trigger its initiation. Intriguingly, memorization prevention mechanisms transfer robustly across different text domains, while memorization induction appears more context-dependent.
Authors: Minjia Mao, Dongjun Wei, Xiao Fang, Michael Chau
Abstract: The proliferation of large language models (LLMs) has significantly transformed the digital information landscape, making it increasingly challenging to distinguish between human-written and LLM-generated content. Detecting LLM-generated information is essential for preserving trust on digital platforms (e.g., social media and e-commerce sites) and preventing the spread of misinformation, a topic that has garnered significant attention in IS research. However, current detection methods, which primarily focus on identifying content generated by specific LLMs in known domains, face challenges in generalizing to new (i.e., unseen) LLMs and domains. This limitation reduces their effectiveness in real-world applications, where the number of LLMs is rapidly multiplying and content spans a vast array of domains. In response, we introduce a general LLM detector (GLD) that combines a twin memory networks design and a theory-guided detection generalization module to detect LLM-generated information across unseen LLMs and domains. Using real-world datasets, we conduct extensive empirical evaluations and case studies to demonstrate the superiority of GLD over state-of-the-art detection methods. The study has important academic and practical implications for digital platforms and LLMs.
Authors: Junqi Jiang, Tom Bewley, Salim I. Amoukou, Francesco Leofante, Antonio Rago, Saumitra Mishra, Francesca Toni
Abstract: Test-time scaling improves large language models' (LLMs) performance by allocating more compute budget during inference. To achieve this, existing methods often require intricate modifications to prompting and sampling strategies. In this work, we introduce representation consistency (RC), a test-time scaling method for aggregating answers drawn from multiple candidate responses of an LLM regardless of how they were generated, including variations in prompt phrasing and sampling strategy. RC enhances answer aggregation by not only considering the number of occurrences of each answer in the candidate response set, but also the consistency of the model's internal activations while generating the set of responses leading to each answer. These activations can be either dense (raw model activations) or sparse (encoded via pretrained sparse autoencoders). Our rationale is that if the model's representations of multiple responses converging on the same answer are highly variable, this answer is more likely to be the result of incoherent reasoning and should be down-weighted during aggregation. Importantly, our method only uses cached activations and lightweight similarity computations and requires no additional model queries. Through experiments with four open-source LLMs and four reasoning datasets, we validate the effectiveness of RC for improving task performance during inference, with consistent accuracy improvements (up to 4%) over strong test-time scaling baselines. We also show that consistency in the sparse activation signals aligns well with the common notion of coherent reasoning.
Authors: Shaoyu Dou, Yutian Shen, Mofan Chen, Zixuan Wang, Jiajie Xu, Qi Guo, Kailai Shao, Chao Chen, Haixiang Hu, Haibo Shi, Min Min, Liwen Zhang
Abstract: Large Language Models (LLMs) demonstrate significant potential but face challenges in complex financial reasoning tasks requiring both domain knowledge and sophisticated reasoning. Current evaluation benchmarks often fall short by not decoupling these capabilities indicators from single task performance and lack root cause analysis for task failure. To address this, we introduce FinEval-KR, a novel evaluation framework for decoupling and quantifying LLMs' knowledge and reasoning abilities independently, proposing distinct knowledge score and reasoning score metrics. Inspired by cognitive science, we further propose a cognitive score based on Bloom's taxonomy to analyze capabilities in reasoning tasks across different cognitive levels. We also release a new open-source Chinese financial reasoning dataset covering 22 subfields to support reproducible research and further advancements in financial reasoning. Our experimental results reveal that LLM reasoning ability and higher-order cognitive ability are the core factors influencing reasoning accuracy. We also specifically find that even top models still face a bottleneck with knowledge application. Furthermore, our analysis shows that specialized financial LLMs generally lag behind the top general large models across multiple metrics.
Authors: Tinh Nguyen, Minh Khue Phan Tran
Abstract: Sign language recognition is crucial for individuals with hearing impairments to break communication barriers. However, previous approaches have had to choose between efficiency and accuracy. Such as RNNs, LSTMs, and GCNs, had problems with vanishing gradients and high computational costs. Despite improving performance, transformer-based methods were not commonly used. This study presents a new novel SLR approach that overcomes the challenge of independently extracting meaningful information from the x and y coordinates of skeleton sequences, which traditional models often treat as inseparable. By utilizing an encoder-decoder of BART architecture, the model independently encodes the x and y coordinates, while Cross-Attention ensures their interrelation is maintained. With only 749,888 parameters, the model achieves 96.04% accuracy on the LSA-64 dataset, significantly outperforming previous models with over one million parameters. The model also demonstrates excellent performance and generalization across WLASL and ASL-Citizen datasets. Ablation studies underscore the importance of coordinate projection, normalization, and using multiple skeleton components for boosting model efficacy. This study offers a reliable and effective approach for sign language recognition, with strong potential for enhancing accessibility tools for the deaf and hard of hearing.
Authors: Ahmed M. Adly, Mostafa Samy, Amr Fawzy
Abstract: We present Gazal-R1, a 32-billion-parameter language model that achieves state-of-the-art performance in medical reasoning while providing transparent, step-by-step explanations for clinical decision-making. Built upon Qwen3 32B, our model demonstrates that strategic training can enable mid-sized models to outperform significantly larger counterparts in specialized domains. We developed a novel two-stage training pipeline: first, supervised fine-tuning on a carefully curated dataset of 107,033 synthetic medical reasoning examples that teaches structured clinical thinking, enhanced by advanced parameter-efficient techniques including Weight-Decomposed Low-Rank Adaptation (DoRA) and Rank-Stabilized LoRA (rsLoRA); second, reinforcement learning using Group Relative Policy Optimization (GRPO) with a sophisticated multi-component reward system that refines accuracy, format adherence, and reasoning quality. Gazal-R1 achieves exceptional performance across medical benchmarks, scoring 87.1% on MedQA, 81.6% on MMLU Pro (Medical), and 79.6% on PubMedQA, surpassing models up to 12x larger. Beyond its strong empirical results, this work provides detailed insights into the challenges of training reasoning-capable models in specialized domains, including issues with reward hacking, training instability, and the fundamental tension between factual recall and detailed reasoning. Our methodology offers a reproducible framework for developing high-capability, domain-specific language models that balance performance, efficiency, and explainability.
Authors: Jinpyo Kim, Gyeongje Cho, Chanwoo Park, Jongwon Park, Jongmin Kim, Yeonkyoun So, Jaejin Lee
Abstract: Since state-of-the-art LLMs often underperform in languages other than English or Chinese, improving the capability of LLMs in new languages has become an essential task. Moreover, LLMs' entire end-to-end training process remains largely unknown to the public due to proprietary reasons, technical complexity, inconsistent documentation, and ethical considerations. The complete picture remains a closely guarded secret within the industry. This paper presents methods to adapt an existing English-based LLM to Korean in a low-budget scenario. We describe the entire end-to-end process: collecting Korean datasets, preprocessing the data, training the model, creating downstream benchmarks, and conducting evaluations. The evaluation results indicate that our method can effectively and cost-efficiently add new language capabilities to existing LLMs. Our new bilingual models, Thunder-LLM and Thunder-LLM-Ins, achieve superior Korean performance compared to state-of-the-art models while utilizing minimal data and computational resources. We share our comprehensive experience and make the code publicly available.
Authors: Hessa A. Alawwad, Anas Zafar, Areej Alhothali, Usman Naseem, Ali Alkhathlan, Amani Jamal
Abstract: Multimodal large language models (MLLMs) have recently achieved significant success in vision--language tasks. However, their capacity to reason over complex, long lessons and intricate educational diagrams that cannot be represented as a single natural image remains largely untested. In this work, we present the first evaluation of state-of-the-art MLLMs on the textbook question answering (TQA) task using the CK12-QA dataset. We assess the performance of recent vision-language models, including LLaVA and LLaMA 3.2-Vision, across various input configurations. Additionally, we introduce a lightweight multimodal retrieval-augmented generation (RAG) pipeline that integrates both paragraphs and diagrams from the lesson into the prompt. Our results demonstrate the influence of retrieved educational context on model accuracy and reasoning, while also revealing current limitations in handling question-context relationships and the potential for noise, pointing to key directions for future research in multimodal AI-driven learning.
Authors: Brandon Colelough, Davis Bartels, Dina Demner-Fushman
Abstract: In this paper, we present an overview of ClinIQLink, a shared task, collocated with the 24th BioNLP workshop at ACL 2025, designed to stress-test large language models (LLMs) on medically-oriented question answering aimed at the level of a General Practitioner. The challenge supplies 4,978 expert-verified, medical source-grounded question-answer pairs that cover seven formats: true/false, multiple choice, unordered list, short answer, short-inverse, multi-hop, and multi-hop-inverse. Participating systems, bundled in Docker or Apptainer images, are executed on the CodaBench platform or the University of Maryland's Zaratan cluster. An automated harness (Task 1) scores closed-ended items by exact match and open-ended items with a three-tier embedding metric. A subsequent physician panel (Task 2) audits the top model responses.
Authors: Chang Liu, Hongkai Chen, Yujun Cai, Hang Wu, Qingwen Ye, Ming-Hsuan Yang, Yiwei Wang
Abstract: Document understanding remains a significant challenge for multimodal large language models (MLLMs). While previous research has primarily focused on locating evidence pages through precise multimodal queries, our work investigates a fundamental yet overlooked aspect: how input format influences document comprehension performance. Through systematic analysis, we discover that raw OCR text often impairs rather than improves MLLMs' performance, which is a counterintuitive finding we attribute to attention dispersion and structure loss. To further substantiate our hypothesis, we propose a novel structure-preserving approach that encodes document elements using the LaTex paradigm, maintaining the hierarchical organization and spatial relationships critical for comprehension. Our attention analysis reveals that structured text induces structured attention patterns on both textual and visual content, directing models to focus on semantically meaningful regions while reducing attention waste. This approach significantly enhances MLLMs' document question answering performance across diverse document types without requiring architectural modifications or additional training.
Authors: Xiaoyan Feng, He Zhang, Yanjun Zhang, Leo Yu Zhang, Shirui Pan
Abstract: Recent advances in Large Language Models (LLMs) have raised urgent concerns about LLM-generated text authenticity, prompting regulatory demands for reliable identification mechanisms. Although watermarking offers a promising solution, existing approaches struggle to simultaneously achieve three critical requirements: text quality preservation, model-agnostic detection, and message embedding capacity, which are crucial for practical implementation. To achieve these goals, the key challenge lies in balancing the trade-off between text quality preservation and message embedding capacity. To address this challenge, we propose BiMark, a novel watermarking framework that achieves these requirements through three key innovations: (1) a bit-flip unbiased reweighting mechanism enabling model-agnostic detection, (2) a multilayer architecture enhancing detectability without compromising generation quality, and (3) an information encoding approach supporting multi-bit watermarking. Through theoretical analysis and extensive experiments, we validate that, compared to state-of-the-art multi-bit watermarking methods, BiMark achieves up to 30% higher extraction rates for short texts while maintaining text quality indicated by lower perplexity, and performs comparably to non-watermarked text on downstream tasks such as summarization and translation.
Authors: Yenisel Plasencia-Cala\~na
Abstract: This paper explores the human-centric operationalization of Automated Essay Scoring (AES) systems, addressing aspects beyond accuracy. We compare various machine learning-based approaches with Large Language Models (LLMs) approaches, identifying their strengths, similarities and differences. The study investigates key dimensions such as bias, robustness, and explainability, considered important for human-aware operationalization of AES systems. Our study shows that ML-based AES models outperform LLMs in accuracy but struggle with explainability, whereas LLMs provide richer explanations. We also found that both approaches struggle with bias and robustness to edge scores. By analyzing these dimensions, the paper aims to identify challenges and trade-offs between different methods, contributing to more reliable and trustworthy AES methods.
Authors: Haoran Tan, Zeyu Zhang, Chen Ma, Xu Chen, Quanyu Dai, Zhenhua Dong
Abstract: Recent works have highlighted the significance of memory mechanisms in LLM-based agents, which enable them to store observed information and adapt to dynamic environments. However, evaluating their memory capabilities still remains challenges. Previous evaluations are commonly limited by the diversity of memory levels and interactive scenarios. They also lack comprehensive metrics to reflect the memory capabilities from multiple aspects. To address these problems, in this paper, we construct a more comprehensive dataset and benchmark to evaluate the memory capability of LLM-based agents. Our dataset incorporates factual memory and reflective memory as different levels, and proposes participation and observation as various interactive scenarios. Based on our dataset, we present a benchmark, named MemBench, to evaluate the memory capability of LLM-based agents from multiple aspects, including their effectiveness, efficiency, and capacity. To benefit the research community, we release our dataset and project at https://github.com/import-myself/Membench.
Authors: Parham Pourdavood, Michael Jacob, Terrence Deacon
Abstract: This paper proposes a novel conceptualization of Large Language Models (LLMs) as externalized informational substrates that function analogously to DNA for human cultural dynamics. Rather than viewing LLMs as either autonomous intelligence or mere programmed mimicry, we argue they serve a broader role as repositories that preserve compressed patterns of human symbolic expression--"fossils" of meaningful dynamics that retain relational residues without their original living contexts. Crucially, these compressed patterns only become meaningful through human reinterpretation, creating a recursive feedback loop where they can be recombined and cycle back to ultimately catalyze human creative processes. Through analysis of four universal features--compression, decompression, externalization, and recursion--we demonstrate that just as DNA emerged as a compressed and externalized medium for preserving useful cellular dynamics without containing explicit reference to goal-directed physical processes, LLMs preserve useful regularities of human culture without containing understanding of embodied human experience. Therefore, we argue that LLMs' significance lies not in rivaling human intelligence, but in providing humanity a tool for self-reflection and playful hypothesis-generation in a low-stakes, simulated environment. This framework positions LLMs as tools for cultural evolvability, enabling humanity to generate novel hypotheses about itself while maintaining the human interpretation necessary to ground these hypotheses in ongoing human aesthetics and norms.
Authors: Dipak Meher, Carlotta Domeniconi, Guadalupe Correa-Cabrera
Abstract: Human smuggling networks are increasingly adaptive and difficult to analyze. Legal case documents offer valuable insights but are unstructured, lexically dense, and filled with ambiguous or shifting references-posing challenges for automated knowledge graph (KG) construction. Existing KG methods often rely on static templates and lack coreference resolution, while recent LLM-based approaches frequently produce noisy, fragmented graphs due to hallucinations, and duplicate nodes caused by a lack of guided extraction. We propose CORE-KG, a modular framework for building interpretable KGs from legal texts. It uses a two-step pipeline: (1) type-aware coreference resolution via sequential, structured LLM prompts, and (2) entity and relationship extraction using domain-guided instructions, built on an adapted GraphRAG framework. CORE-KG reduces node duplication by 33.28%, and legal noise by 38.37% compared to a GraphRAG-based baseline-resulting in cleaner and more coherent graph structures. These improvements make CORE-KG a strong foundation for analyzing complex criminal networks.
Authors: Yasmine Bouamra, Bruno Yun, Alexandre Poisson, Fr\'ed\'eric Armetta
Abstract: The automatic generation of SysML v2 models represents a major challenge in the engineering of complex systems, particularly due to the scarcity of learning corpora and complex syntax. We present SysTemp, a system aimed at facilitating and improving the creation of SysML v2 models from natural language specifications. It is based on a multi-agent system, including a template generator that structures the generation process. We discuss the advantages and challenges of this system through an evaluation, highlighting its potential to improve the quality of the generations in SysML v2 modeling.
Authors: Junhao Liu, Zhenhao Xu, Yuxin Fang, Yichuan Chen, Zuobin Ying, Wenhan Chang
Abstract: Recently, there have been notable advancements in large language models (LLMs), demonstrating their growing abilities in complex reasoning. However, existing research largely overlooks a thorough and systematic comparison of these models' reasoning processes and outputs, particularly regarding their self-reflection pattern (also termed "Aha moment") and the interconnections across diverse domains. This paper proposes a novel framework for analyzing the reasoning characteristics of four cutting-edge large reasoning models (GPT-o1, DeepSeek-R1, Kimi-k1.5, and Grok-3) using keywords statistic and LLM-as-a-judge paradigm. Our approach connects their internal thinking processes with their final outputs. A diverse dataset consists of real-world scenario-based questions covering logical deduction, causal inference, and multi-step problem-solving. Additionally, a set of metrics is put forward to assess both the coherence of reasoning and the accuracy of the outputs. The research results uncover various patterns of how these models balance exploration and exploitation, deal with problems, and reach conclusions during the reasoning process. Through quantitative and qualitative comparisons, disparities among these models are identified in aspects such as the depth of reasoning, the reliance on intermediate steps, and the degree of similarity between their thinking processes and output patterns and those of GPT-o1. This work offers valuable insights into the trade-off between computational efficiency and reasoning robustness and provides practical recommendations for enhancing model design and evaluation in practical applications. We publicly release our project at: https://github.com/ChangWenhan/FromThinking2Output
Authors: Xiyuan Zhang, Boran Han, Haoyang Fang, Abdul Fatir Ansari, Shuai Zhang, Danielle C. Maddix, Cuixiong Hu, Andrew Gordon Wilson, Michael W. Mahoney, Hao Wang, Yan Liu, Huzefa Rangwala, George Karypis, Bernie Wang
Abstract: Recently, there has been growing interest in incorporating textual information into foundation models for time series forecasting. However, it remains unclear whether and under what conditions such multimodal integration consistently yields gains. We systematically investigate these questions across a diverse benchmark of 14 forecasting tasks spanning 7 domains, including health, environment, and economics. We evaluate two popular multimodal forecasting paradigms: aligning-based methods, which align time series and text representations; and prompting-based methods, which directly prompt large language models for forecasting. Although prior works report gains from multimodal input, we find these effects are not universal across datasets and models, and multimodal methods sometimes do not outperform the strongest unimodal baselines. To understand when textual information helps, we disentangle the effects of model architectural properties and data characteristics. Our findings highlight that on the modeling side, incorporating text information is most helpful given (1) high-capacity text models, (2) comparatively weaker time series models, and (3) appropriate aligning strategies. On the data side, performance gains are more likely when (4) sufficient training data is available and (5) the text offers complementary predictive signal beyond what is already captured from the time series alone. Our empirical findings offer practical guidelines for when multimodality can be expected to aid forecasting tasks, and when it does not.
Authors: Xiaobin Ren, Xinyu Zhu, Kaiqi Zhao
Abstract: Currently, considerable strides have been achieved in Point-of-Interest (POI) embedding methodologies, driven by the emergence of novel POI tasks like recommendation and classification. Despite the success of task-specific, end-to-end models in POI embedding, several challenges remain. These include the need for more effective multi-context sampling strategies, insufficient exploration of multiple POI contexts, limited versatility, and inadequate generalization. To address these issues, we propose the AdaptGOT model, which integrates both the (Adapt)ive representation learning technique and the Geographical-Co-Occurrence-Text (GOT) representation with a particular emphasis on Geographical location, Co-Occurrence and Textual information. The AdaptGOT model comprises three key components: (1) contextual neighborhood generation, which integrates advanced mixed sampling techniques such as KNN, density-based, importance-based, and category-aware strategies to capture complex contextual neighborhoods; (2) an advanced GOT representation enhanced by an attention mechanism, designed to derive high-quality, customized representations and efficiently capture complex interrelations between POIs; and (3) the MoE-based adaptive encoder-decoder architecture, which ensures topological consistency and enriches contextual representation by minimizing Jensen-Shannon divergence across varying contexts. Experiments on two real-world datasets and multiple POI tasks substantiate the superior performance of the proposed AdaptGOT model.
Authors: Gautam Siddharth Kashyap, Mohammad Anas Azeez, Rafiq Ali, Zohaib Hasan Siddiqui, Jiechao Gao, Usman Naseem
Abstract: The increasing prevalence of child-targeted hate speech online underscores the urgent need for specialized datasets to address this critical issue. Existing hate speech datasets lack agespecific annotations, fail to capture nuanced contexts, and overlook the unique emotional impact on children. To bridge this gap, we introduce ChildGuard1, a curated dataset derived from existing corpora and enriched with child-specific annotations. ChildGuard captures diverse contexts of child-targeted hate speech, spanning age groups. We benchmark existing state-of-the-art hate speech detection methods, including Large Language Models (LLMs), and assess their effectiveness in detecting and contextualizing child-targeted hate speech. To foster further research in this area, we publicly release ChildGuard, providing a robust foundation for developing improved methods to detect and mitigate such harm.
Authors: Yixiong Fang, Tianran Sun, Yuling Shi, Min Wang, Xiaodong Gu
Abstract: The increasing complexity of large language models (LLMs) raises concerns about their ability to "cheat" on standard Question Answering (QA) benchmarks by memorizing task-specific data. This undermines the validity of benchmark evaluations, as they no longer reflect genuine model capabilities but instead the effects of data leakage. While prior work has focused on detecting such leakage, little attention has been given to mitigating its impact and preserving the long-term utility of benchmarks. In this paper, we introduce LastingBench, a novel framework designed to continuously reinforce and safeguard existing benchmarks against knowledge leakage. LastingBench identifies leakage points in the context through perturbation, then rewrites the leakage points to counterfactual ones-disrupting memorization while preserving the benchmark's original evaluative intent. Evaluations of state-of-the-art QA benchmarks show significant performance gaps, highlighting the efficacy of LastingBench in reducing memorization effects. LastingBench offers a practical and scalable solution to ensure benchmark robustness over time, promoting fairer and more interpretable evaluations of LLMs.
Authors: Wenhao Li, Hongkuan Zhang, Hongwei Zhang, Zhengxu Li, Zengjie Dong, Yafan Chen, Niranjan Bidargaddi, Hong Liu
Abstract: Current medical language models, adapted from large language models (LLMs), typically predict ICD code-based diagnosis from electronic health records (EHRs) because these labels are readily available. However, ICD codes do not capture the nuanced, context-rich reasoning clinicians use for diagnosis. Clinicians synthesize diverse patient data and reference clinical practice guidelines (CPGs) to make evidence-based decisions. This misalignment limits the clinical utility of existing models. We introduce GARMLE-G, a Generation-Augmented Retrieval framework that grounds medical language model outputs in authoritative CPGs. Unlike conventional Retrieval-Augmented Generation based approaches, GARMLE-G enables hallucination-free outputs by directly retrieving authoritative guideline content without relying on model-generated text. It (1) integrates LLM predictions with EHR data to create semantically rich queries, (2) retrieves relevant CPG knowledge snippets via embedding similarity, and (3) fuses guideline content with model output to generate clinically aligned recommendations. A prototype system for hypertension diagnosis was developed and evaluated on multiple metrics, demonstrating superior retrieval precision, semantic relevance, and clinical guideline adherence compared to RAG-based baselines, while maintaining a lightweight architecture suitable for localized healthcare deployment. This work provides a scalable, low-cost, and hallucination-free method for grounding medical language models in evidence-based clinical practice, with strong potential for broader clinical deployment.
Authors: Chuanrui Hu, Wei Hu, Penghang Yu, Hua Zhang, Bing-Kun Bao
Abstract: Open-domain Timeline Summarization (TLS) is crucial for monitoring the evolution of news topics. To identify changes in news topics, existing methods typically employ general Large Language Models (LLMs) to summarize relevant timestamps from retrieved news. While general LLMs demonstrate capabilities in zero-shot news summarization and timestamp localization, they struggle with assessing topic relevance and understanding topic evolution. Consequently, the summarized information often includes irrelevant details or inaccurate timestamps. To address these issues, we propose the first large Timeline Intelligence Model (TIM) for open-domain TLS, which is capable of effectively summarizing open-domain timelines. Specifically, we begin by presenting a large-scale TLS dataset, comprising over 1,000 news topics and more than 3,000 annotated TLS instances. Furthermore, we propose a progressive optimization strategy, which gradually enhance summarization performance. It employs instruction tuning to enhance summarization and topic-irrelevant information filtering capabilities. Following this, it exploits a novel dual-alignment reward learning method that incorporates both semantic and temporal perspectives, thereby improving the understanding of topic evolution principles. Through this progressive optimization strategy, TIM demonstrates a robust ability to summarize open-domain timelines. Extensive experiments in open-domain demonstrate the effectiveness of our TIM.
Authors: Zhiyuan Zhang, Xiaosong Jia, Guanyu Chen, Qifeng Li, Junchi Yan
Abstract: In this technical report, we introduce TrajTok, a trajectory tokenizer for discrete next-token-prediction based behavior generation models, which combines data-driven and rule-based methods with better coverage, symmetry and robustness, along with a spatial-aware label smoothing method for cross-entropy loss. We adopt the tokenizer and loss for the SMART model and reach a superior performance with realism score of 0.7852 on the Waymo Open Sim Agents Challenge 2025. We will open-source the code in the future.
Authors: Siyi Zhou, Yiquan Zhou, Yi He, Xun Zhou, Jinchao Wang, Wei Deng, Jingchen Shu
Abstract: Large-scale text-to-speech (TTS) models are typically categorized into autoregressive and non-autoregressive systems. Although autoregressive systems exhibit certain advantages in speech naturalness, their token-by-token generation mechanism makes it difficult to precisely control the duration of synthesized speech. This is a key limitation in applications such as video dubbing that require strict audio-visual synchronization. This paper introduces IndexTTS2, which proposes a novel and autoregressive-model-friendly method for speech duration control. The method supports two generation modes: one allows explicit specification of the number of generated tokens for precise duration control; the other does not require manual input and lets the model freely generate speech while preserving prosodic characteristics from the input prompt. Furthermore, IndexTTS2 achieves disentanglement between emotional expression and speaker identity, enabling independent control of timbre and emotion. In the zero-shot setting, the model can perfectly reproduce the emotional characteristics of the input prompt. Users may also provide a separate emotion prompt, even from a different speaker, allowing the model to reconstruct the target timbre while conveying the desired emotion. To enhance clarity during strong emotional expressions, we incorporate GPT latent representations to improve speech stability. Meanwhile, to lower the barrier for emotion control, we design a soft instruction mechanism based on textual descriptions by fine-tuning Qwen3. This enables effective guidance of speech generation with desired emotional tendencies using natural language input. Experimental results demonstrate that IndexTTS2 outperforms existing state-of-the-art zero-shot TTS models in word error rate, speaker similarity, and emotional fidelity.
Authors: Daniele Cirulli, Giulio Cimini, Giovanni Palermo
Abstract: Large Language Models (LLMs) have recently emerged as powerful tools for natural language generation, with applications spanning from content creation to social simulations. Their ability to mimic human interactions raises both opportunities and concerns, particularly in the context of politically relevant online discussions. In this study, we evaluate the performance of LLMs in replicating user-generated content within a real-world, divisive scenario: Reddit conversations during the 2016 US Presidential election. In particular, we conduct three different experiments, asking GPT-4 to generate comments by impersonating either real or artificial partisan users. We analyze the generated comments in terms of political alignment, sentiment, and linguistic features, comparing them against real user contributions and benchmarking against a null model. We find that GPT-4 is able to produce realistic comments, both in favor of or against the candidate supported by the community, yet tending to create consensus more easily than dissent. In addition we show that real and artificial comments are well separated in a semantically embedded space, although they are indistinguishable by manual inspection. Our findings provide insights on the potential use of LLMs to sneak into online discussions, influence political debate and shape political narratives, bearing broader implications of AI-driven discourse manipulation.
Authors: Jasper Dekoninck, Ivo Petrov, Kristian Minchev, Mislav Balunovic, Martin Vechev, Miroslav Marinov, Maria Drencheva, Lyuba Konova, Milen Shumanov, Kaloyan Tsvetkov, Nikolay Drenchev, Lazar Todorov, Kalina Nikolova, Nikolay Georgiev, Vanesa Kalinkova, Margulan Ismoldayev
Abstract: In recent months, large language models (LLMs) have made significant progress in mathematical proof generation, but further advancement is hindered by the lack of a large-scale, high-quality dataset of human-evaluated proofs. While expensive to create, such a dataset is essential for driving improvements in training and enabling a rigorous analysis of proof generation capabilities. In this work, we present the Open Proof Corpus (OPC), a dataset comprising over 5,000 human-evaluated proofs produced by state-of-the-art LLMs. The OPC was specifically designed for broad applicability and downstream usage in proof generation research and is the first to include a substantial number of correct, LLM-generated solutions to problems from prestigious mathematics competitions such as the USAMO and IMO. Using the OPC, we explore critical questions in automated proof generation: (1) the performance gap between natural language and formal proof generation, (2) the discrepancy between final-answer accuracy and full-proof validity, and (3) the impact of best-of-n selection on proof quality. Finally, to showcase the utility of the OPC, we finetune an 8B-parameter model on the dataset, obtaining a model that performs on par with the best model, Gemini-2.5-Pro, on the task of evaluating proof correctness.
Authors: Niclas Pokel, Pehu\'en Moure, Roman Boehringer, Yingqiang Gao
Abstract: Speech impairments caused by conditions such as cerebral palsy or genetic disorders pose significant challenges for automatic speech recognition (ASR) systems. Despite recent advances, ASR models like Whisper struggle with non-normative speech due to limited training data and the difficulty of collecting and annotating non-normative speech samples. In this work, we propose a practical and lightweight pipeline to personalize ASR models, formalizing the selection of words and enriching a small, speech-impaired dataset with semantic coherence. Applied to data from a child with a structural speech impairment, our approach shows promising improvements in transcription quality, demonstrating the potential to reduce communication barriers for individuals with atypical speech patterns.
Authors: Peiheng Gao, Chen Yang, Ning Sun, Ri\v{c}ardas Zitikis
Abstract: Machine learning (ML) has significantly advanced text classification by enabling automated understanding and categorization of complex, unstructured textual data. However, accurately capturing nuanced linguistic patterns and contextual variations inherent in natural language, particularly within consumer complaints, remains a challenge. This study addresses these issues by incorporating human-experience-trained algorithms that effectively recognize subtle semantic differences crucial for assessing consumer relief eligibility. Furthermore, we propose integrating synthetic data generation methods that utilize expert evaluations of generative adversarial networks and are refined through expert annotations. By combining expert-trained classifiers with high-quality synthetic data, our research seeks to significantly enhance machine learning classifier performance, reduce dataset acquisition costs, and improve overall evaluation metrics and robustness in text classification tasks.
Authors: Jiaxi Zhuang, Kangning Li, Jue Hou, Mingjun Xu, Zhifeng Gao, Hengxing Cai
Abstract: Extracting molecular structure-activity relationships (SARs) from scientific literature and patents is essential for drug discovery and materials research. However, this task remains challenging due to heterogeneous document formats and limitations of existing methods. Specifically, rule-based approaches relying on rigid templates fail to generalize across diverse document layouts, while general-purpose multimodal large language models (MLLMs) lack sufficient accuracy and reliability for specialized tasks, such as layout detection and optical chemical structure recognition (OCSR). To address these challenges, we introduce DocSAR-200, a rigorously annotated benchmark of 200 scientific documents designed specifically for evaluating SAR extraction methods. Additionally, we propose Doc2SAR, a novel synergistic framework that integrates domain-specific tools with MLLMs enhanced via supervised fine-tuning (SFT). Extensive experiments demonstrate that Doc2SAR achieves state-of-the-art performance across various document types, significantly outperforming leading end-to-end baselines. Specifically, Doc2SAR attains an overall Table Recall of 80.78% on DocSAR-200, exceeding end2end GPT-4o by 51.48%. Furthermore, Doc2SAR demonstrates practical usability through efficient inference and is accompanied by a web app.
Authors: Li Zhou, Hao Jiang, Junjie Li, Zefeng Zhao, Feng Jiang, Wenyu Chen, Haizhou Li
Abstract: Explicit structural information has been proven to be encoded by Graph Neural Networks (GNNs), serving as auxiliary knowledge to enhance model capabilities and improve performance in downstream NLP tasks. However, recent studies indicate that GNNs fail to fully utilize structural information, whereas Multi-Layer Perceptrons (MLPs), despite lacking the message-passing mechanisms inherent to GNNs, exhibit a surprising ability in structure-aware tasks. Motivated by these findings, this paper introduces a comprehensive probing framework from an information-theoretic perspective. The framework is designed to systematically assess the role of explicit structural modeling in enhancing language model (LM) representations and to investigate the potential of MLPs as efficient and scalable alternatives to GNNs. We extend traditional probing classifiers by incorporating a control module that allows for selective use of either the full GNN model or its decoupled components, specifically, the message-passing and feature-transformation operations.This modular approach isolates and assesses the individual contributions of these operations, avoiding confounding effects from the complete GNN architecture. Using the Edge Probing Suite, a diagnostic tool for evaluating the linguistic knowledge encoded in LMs, we find that MLPs, when used as feature-transformation modules, consistently improve the linguistic knowledge captured in LM representations across different architectures. They effectively encode both syntactic and semantic patterns. Similarly, GNNs that incorporate feature-transformation operations show beneficial effects. In contrast, models that rely solely on message-passing operations tend to underperform, often leading to negative impacts on probing task performance.
Authors: Swastika Kundu, Autoshi Ibrahim, Mithila Rahman, Tanvir Ahmed
Abstract: Sentiment analysis for regional dialects of Bangla remains an underexplored area due to linguistic diversity and limited annotated data. This paper introduces ANUBHUTI, a comprehensive dataset consisting of 2000 sentences manually translated from standard Bangla into four major regional dialects Mymensingh, Noakhali, Sylhet, and Chittagong. The dataset predominantly features political and religious content, reflecting the contemporary socio political landscape of Bangladesh, alongside neutral texts to maintain balance. Each sentence is annotated using a dual annotation scheme: multiclass thematic labeling categorizes sentences as Political, Religious, or Neutral, and multilabel emotion annotation assigns one or more emotions from Anger, Contempt, Disgust, Enjoyment, Fear, Sadness, and Surprise. Expert native translators conducted the translation and annotation, with quality assurance performed via Cohens Kappa inter annotator agreement, achieving strong consistency across dialects. The dataset was further refined through systematic checks for missing data, anomalies, and inconsistencies. ANUBHUTI fills a critical gap in resources for sentiment analysis in low resource Bangla dialects, enabling more accurate and context aware natural language processing.
Authors: Tzu-Quan Lin, Hsi-Chun Cheng, Hung-yi Lee, Hao Tang
Abstract: In recent years, the impact of self-supervised speech Transformers has extended to speaker-related applications. However, little research has explored how these models encode speaker information. In this work, we address this gap by identifying neurons in the feed-forward layers that are correlated with speaker information. Specifically, we analyze neurons associated with k-means clusters of self-supervised features and i-vectors. Our analysis reveals that these clusters correspond to broad phonetic and gender classes, making them suitable for identifying neurons that represent speakers. By protecting these neurons during pruning, we can significantly preserve performance on speaker-related task, demonstrating their crucial role in encoding speaker information.
Authors: Eivind Morris Bakke, Nora Winger Heggelund
Abstract: Automatic fact verification systems increasingly rely on large language models (LLMs). We investigate how parametric knowledge biases in these models affect fact-checking outcomes of the HerO system (baseline for FEVER-25). We examine how the system is affected by: (1) potential bias in Llama 3.1's parametric knowledge and (2) intentionally injected bias. When prompted directly to perform fact-verification, Llama 3.1 labels nearly half the claims as "Not Enough Evidence". Using only its parametric knowledge it is able to reach a verdict on the remaining half of the claims. In the second experiment, we prompt the model to generate supporting, refuting, or neutral fact-checking documents. These prompts significantly influence retrieval outcomes, with approximately 50\% of retrieved evidence being unique to each perspective. Notably, the model sometimes refuses to generate supporting documents for claims it believes to be false, creating an inherent negative bias. Despite differences in retrieved evidence, final verdict predictions show stability across prompting strategies. The code is available at: https://github.com/eibakke/FEVER-8-Shared-Task
Authors: Alexandru Dumitru, V Venktesh, Adam Jatowt, Avishek Anand
Abstract: Large Language Models (LLMs) have demonstrated immense advances in a wide range of natural language tasks. However, these models are susceptible to hallucinations and errors on particularly temporal understanding tasks involving multiple entities in answers. In such tasks, they fail to associate entities with accurate time intervals, generate a complete list of entities in answers or reason about events associated with specific temporal bounds. Existing works do not extensively evaluate the abilities of the model to perform implicit and explicit temporal understanding in a list answer construction setup. To bridge this gap, we propose the Time referenced List based Question Answering or TLQA benchmark that requires structured answers in list format aligned with corresponding time periods. Our TLQA benchmark, requires both list construction and temporal understanding simultaneously, which to the best of our knowledge has not been explored in prior benchmarks. We investigate the temporal understanding and list construction capabilities of state-of-the-art generative models on TLQA in closed-book and open-domain settings. Our findings reveal significant shortcomings in current models, particularly their inability to provide complete answers and temporally align facts in a closed-book setup and the need to improve retrieval in open-domain setup, providing clear future directions for research on TLQA. The benchmark and code at https://github.com/elixir-research-group/TLQA.
Authors: Reem Alothman, Hafida Benhidour, Said Kerrache
Abstract: The widespread use of text-based communication on social media-through chats, comments, and microblogs-has improved user interaction but has also led to an increase in offensive content, including hate speech, racism, and other forms of abuse. Due to the enormous volume of user-generated content, manual moderation is impractical, which creates a need for automated systems that can detect offensive language. Deep learning models, particularly those using transfer learning, have demonstrated significant success in understanding natural language through large-scale pretraining. In this study, we propose an automatic offensive language detection model based on XLNet, a generalized autoregressive pretraining method, and compare its performance with BERT (Bidirectional Encoder Representations from Transformers), which is a widely used baseline in natural language processing (NLP). Both models are evaluated using the Offensive Language Identification Dataset (OLID), a benchmark Twitter dataset that includes hierarchical annotations. Our experimental results show that XLNet outperforms BERT in detecting offensive content and in categorizing the types of offenses, while BERT performs slightly better in identifying the targets of the offenses. Additionally, we find that oversampling and undersampling strategies are effective in addressing class imbalance and improving classification performance. These findings highlight the potential of transfer learning and XLNet-based architectures to create robust systems for detecting offensive language on social media platforms.
Authors: Jonathan St-Onge, Ashley M. A. Fehr, Carter Ward, Calla G. Beauregard, Michael V. Arnold, Samuel F. Rosenblatt, Benjamin Cooley, Christopher M. Danforth, Peter Sheridan Dodds
Abstract: Describing and comparing complex systems requires principled, theoretically grounded tools. Built around the phenomenon of type turbulence, allotaxonographs provide map-and-list visual comparisons of pairs of heavy-tailed distributions. Allotaxonographs are designed to accommodate a wide range of instruments including rank- and probability-turbulence divergences, Jenson-Shannon divergence, and generalized entropy divergences. Here, we describe a suite of programmatic tools for rendering allotaxonographs for rank-turbulence divergence in Matlab, Javascript, and Python, all of which have different use cases.
Authors: Avash Palikhe, Zhenyu Yu, Zichong Wang, Wenbin Zhang
Abstract: Large Language Models (LLMs) have played a pivotal role in advancing Artificial Intelligence (AI). However, despite their achievements, LLMs often struggle to explain their decision-making processes, making them a 'black box' and presenting a substantial challenge to explainability. This lack of transparency poses a significant obstacle to the adoption of LLMs in high-stakes domain applications, where interpretability is particularly essential. To overcome these limitations, researchers have developed various explainable artificial intelligence (XAI) methods that provide human-interpretable explanations for LLMs. However, a systematic understanding of these methods remains limited. To address this gap, this survey provides a comprehensive review of explainability techniques by categorizing XAI methods based on the underlying transformer architectures of LLMs: encoder-only, decoder-only, and encoder-decoder models. Then these techniques are examined in terms of their evaluation for assessing explainability, and the survey further explores how these explanations are leveraged in practical applications. Finally, it discusses available resources, ongoing research challenges, and future directions, aiming to guide continued efforts toward developing transparent and responsible LLMs.
Authors: Riley Galpin, Bryce Anderson, Tom S. Juzek
Abstract: Scientific English has undergone rapid and unprecedented changes in recent years, with words such as "delve," "intricate," and "crucial" showing significant spikes in frequency since around 2022. These changes are widely attributed to the growing influence of Large Language Models like ChatGPT in the discourse surrounding bias and misalignment. However, apart from changes in frequency, the exact structure of these linguistic shifts has remained unclear. The present study addresses this and investigates whether these changes involve the replacement of synonyms by suddenly 'spiking words,' for example, "crucial" replacing "essential" and "key," or whether they reflect broader semantic and pragmatic qualifications. To further investigate structural changes, we include part of speech tagging in our analysis to quantify linguistic shifts over grammatical categories and differentiate between word forms, like "potential" as a noun vs. as an adjective. We systematically analyze synonym groups for widely discussed 'spiking words' based on frequency trends in scientific abstracts from PubMed. We find that entire semantic clusters often shift together, with most or all words in a group increasing in usage. This pattern suggests that changes induced by Large Language Models are primarily semantic and pragmatic rather than purely lexical. Notably, the adjective "important" shows a significant decline, which prompted us to systematically analyze decreasing lexical items. Our analysis of "collapsing" words reveals a more complex picture, which is consistent with organic language change and contrasts with the patterns of the abrupt spikes. These insights into the structure of language change contribute to our understanding of how language technology continues to shape human language.
Authors: Kourosh Shahnazari, Mohammadali Keshtparvar, Seyed Moein Ayyoubzadeh
Abstract: The intricate linguistic, stylistic, and metrical aspects of Persian classical poetry pose a challenge for computational authorship attribution. In this work, we present a versatile framework to determine authorship among 67 prominent poets. We employ a multi-input neural framework consisting of a transformer-based language encoder complemented by features addressing the semantic, stylometric, and metrical dimensions of Persian poetry. Our feature set encompasses 100-dimensional Word2Vec embeddings, seven stylometric measures, and categorical encodings of poetic form and meter. We compiled a vast corpus of 647,653 verses of the Ganjoor digital collection, validating the data through strict preprocessing and author verification while preserving poem-level splitting to prevent overlap. This work employs verse-level classification and majority and weighted voting schemes in evaluation, revealing that weighted voting yields 71% accuracy. We further investigate threshold-based decision filtering, allowing the model to generate highly confident predictions, achieving 97% accuracy at a 0.9 threshold, though at lower coverage. Our work focuses on the integration of deep representational forms with domain-specific features for improved authorship attribution. The results illustrate the potential of our approach for automated classification and the contribution to stylistic analysis, authorship disputes, and general computational literature research. This research will facilitate further research on multilingual author attribution, style shift, and generative modeling of Persian poetry.
Authors: Duo Zhang, Junyi Mo
Abstract: Deep learning has significantly advanced NLP, but its reliance on large black-box models introduces critical interpretability and computational efficiency concerns. This paper proposes LinguaSynth, a novel text classification framework that strategically integrates five complementary linguistic feature types: lexical, syntactic, entity-level, word-level semantics, and document-level semantics within a transparent logistic regression model. Unlike transformer-based architectures, LinguaSynth maintains interpretability and computational efficiency, achieving an accuracy of 84.89 percent on the 20 Newsgroups dataset and surpassing a robust TF-IDF baseline by 3.32 percent. Through rigorous feature interaction analysis, we show that syntactic and entity-level signals provide essential disambiguation and effectively complement distributional semantics. LinguaSynth sets a new benchmark for interpretable, resource-efficient NLP models and challenges the prevailing assumption that deep neural networks are necessary for high-performing text classification.
Authors: Quan Xiao, Debarun Bhattacharjya, Balaji Ganesan, Radu Marinescu, Katsiaryna Mirylenka, Nhan H Pham, Michael Glass, Junkyu Lee
Abstract: Estimating the confidence of large language model (LLM) outputs is essential for real-world applications requiring high user trust. Black-box uncertainty quantification (UQ) methods, relying solely on model API access, have gained popularity due to their practical benefits. In this paper, we examine the implicit assumption behind several UQ methods, which use generation consistency as a proxy for confidence, an idea we formalize as the consistency hypothesis. We introduce three mathematical statements with corresponding statistical tests to capture variations of this hypothesis and metrics to evaluate LLM output conformity across tasks. Our empirical investigation, spanning 8 benchmark datasets and 3 tasks (question answering, text summarization, and text-to-SQL), highlights the prevalence of the hypothesis under different settings. Among the statements, we highlight the `Sim-Any' hypothesis as the most actionable, and demonstrate how it can be leveraged by proposing data-free black-box UQ methods that aggregate similarities between generations for confidence estimation. These approaches can outperform the closest baselines, showcasing the practical value of the empirically observed consistency hypothesis.
Authors: Taiga Someya, Ryo Yoshida, Hitomi Yanaka, Yohei Oseki
Abstract: Recent work has demonstrated that neural language models encode syntactic structures in their internal representations, yet the derivations by which these structures are constructed across layers remain poorly understood. In this paper, we propose Derivational Probing to investigate how micro-syntactic structures (e.g., subject noun phrases) and macro-syntactic structures (e.g., the relationship between the root verbs and their direct dependents) are constructed as word embeddings propagate upward across layers. Our experiments on BERT reveal a clear bottom-up derivation: micro-syntactic structures emerge in lower layers and are gradually integrated into a coherent macro-syntactic structure in higher layers. Furthermore, a targeted evaluation on subject-verb number agreement shows that the timing of constructing macro-syntactic structures is critical for downstream performance, suggesting an optimal timing for integrating global syntactic information.
Authors: Hang Shao, Heting Gao, Yunhang Shen, Jiawei Chen, Lijiang Li, Zuwei Long, Bo Tong, Ke Li, Xing Sun
Abstract: Native multimodal large language models (MLLMs) restructure a single large language model (LLM) into a spoken language model (SLM) capable of both speech and text generation. Compared to modular and aligned MLLMs, native MLLMs preserve richer paralinguistic features such as emotion and prosody, and generate speech responses directly within the backbone LLM rather than using a separate speech decoder. This integration also results in lower response latency and smoother interaction. However, native MLLMs suffer from catastrophic forgetting and performance degradation because the available paired speech-text data is insufficient to support the pretraining of MLLMs compared to the vast amount of text data required to pretrain text LLMs. To address this issue, we propose DeepTalk, a framework for adaptive modality expert learning based on a Mixture of Experts (MoE) architecture. DeepTalk first adaptively distinguishes modality experts according to their modality load within the LLM. Each modality expert then undergoes specialized single-modality training, followed by joint multimodal collaborative training. As a result, DeepTalk incurs only a 5.5% performance drop compared to the original LLM, which is significantly lower than the average performance drop of over 20% typically seen in native MLLMs (such as GLM-4-Voice), and is on par with modular MLLMs. Meanwhile, the end-to-end dialogue latency remains within 0.5 seconds, ensuring a seamless and intelligent speech interaction experience. Code and models are released at https://github.com/talkking/DeepTalk.
Authors: Jian Zhang, Linhao Zhang, Bokai Lei, Chuhan Wu, Wei Jia, Xiao Zhou
Abstract: Recent multi-modal Large Language Models (LLMs) such as GPT-4o have demonstrated strong capabilities of direct speech interaction. However, the lack of specialized and comprehensive benchmarks for end-to-end speech LLM evaluation hinders optimizing the user experience of Audio LLMs in real-world applications. Existing evaluation methods often adapt text-based benchmarks, overlooking speech's unique characteristics and challenges, including prosody, homophones, stuttering, and differing user expectations. Here, we present a novel approach to thoroughly evaluate LLMs in practical speech conversations. We systematically curate real-world chat data relevant to spoken scenarios, introduce diversity in speaker attributes and acoustic conditions, and augment the dataset with speech-specific phenomena. We further design a query-aware evaluation method to use customized evaluation checklists and prompts to enhance the accuracy of automatic evaluation. We conduct comprehensive testing and detailed analysis of various mainstream speech models, revealing significant differences in model performance across different speech scenarios. The use of query-aware evaluation further enables a finer-grained assessment under various speech-specific scenarios. Our benchmark can provide valuable insights for speech model development and evaluation.
Authors: Qiyue Gao, Xinyu Pi, Kevin Liu, Junrong Chen, Ruolan Yang, Xinqi Huang, Xinyu Fang, Lu Sun, Gautham Kishore, Bo Ai, Stone Tao, Mengyang Liu, Jiaxi Yang, Chao-Jung Lai, Chuanyang Jin, Jiannan Xiang, Benhao Huang, Zeming Chen, David Danks, Hao Su, Tianmin Shu, Ziqiao Ma, Lianhui Qin, Zhiting Hu
Abstract: Internal world models (WMs) enable agents to understand the world's state and predict transitions, serving as the basis for advanced deliberative reasoning. Recent large Vision-Language Models (VLMs), such as OpenAI o3, GPT-4o and Gemini, exhibit potential as general-purpose WMs. While the latest studies have evaluated and shown limitations in specific capabilities such as visual understanding, a systematic evaluation of VLMs' fundamental WM abilities remains absent. Drawing on comparative psychology and cognitive science, we propose a two-stage framework that assesses Perception (visual, spatial, temporal, quantitative, and motion) and Prediction (mechanistic simulation, transitive inference, compositional inference) to provide an atomic evaluation of VLMs as WMs. Guided by this framework, we introduce WM-ABench, a large-scale benchmark comprising 23 fine-grained evaluation dimensions across 6 diverse simulated environments with controlled counterfactual simulations. Through 660 experiments on 15 latest commercial and open-source VLMs, we find that these models exhibit striking limitations in basic world modeling abilities. For instance, almost all models perform at near-random accuracy when distinguishing motion trajectories. Additionally, they lack disentangled understanding -- e.g., some models tend to believe blue objects move faster than green ones. More rich results and analyses reveal significant gaps between VLMs and human-level world modeling.
Authors: Sean Kim, Hyuhng Joon Kim
Abstract: As large language models (LLMs) are increasingly deployed across diverse linguistic and cultural contexts, understanding their behavior in both factual and disputable scenarios is essential, especially when their outputs may shape public opinion or reinforce dominant narratives. In this paper, we define two types of bias in LLMs: model bias (bias stemming from model training) and inference bias (bias induced by the language of the query), through a two-phase evaluation. Phase 1 evaluates LLMs on factual questions where a single verifiable answer exists, assessing whether models maintain consistency across different query languages. Phase 2 expands the scope by probing geopolitically sensitive disputes, where responses may reflect culturally embedded or ideologically aligned perspectives. We construct a manually curated dataset spanning both factual and disputable QA, across four languages and question types. The results show that Phase 1 exhibits query language induced alignment, while Phase 2 reflects an interplay between the model's training context and query language. This paper offers a structured framework for evaluating LLM behavior across neutral and sensitive topics, providing insights for future LLM deployment and culturally aware evaluation practices in multilingual contexts.
Authors: Ernie Chang, Yang Li, Patrick Huber, David Kant, Yangyang Shi, Vikas Chandra
Abstract: In language model training, it is desirable to equip models with capabilities from various tasks. However, it is not clear how to directly obtain the right data mixtures for these capabilities as the relationship between data and tasks is difficult to be modeled. In this work, we observe that checkpoint models exhibit emerging capabilities at different points in the training trajectory. Often, the training process saves checkpoints as artifacts that are under-utilized as a source of in-training data signals. We identify these artifact models based on their respective capabilities on the benchmarks and leverage them as data mixers by using their aggregated first-order influence approximation over source data. We demonstrated on eight reasoning benchmarks that the proposed framework shows significant improvements in the pretraining setting, with performance improvements of up to 1.93%. Overall, this shows the potential of checkpoint models to enhance data quality and optimize data mixtures.
Authors: Junho Myung, Yeon Su Park, Sunwoo Kim, Shin Yoo, Alice Oh
Abstract: Evaluating the performance and biases of large language models (LLMs) through role-playing scenarios is becoming increasingly common, as LLMs often exhibit biased behaviors in these contexts. Building on this line of research, we introduce PapersPlease, a benchmark consisting of 3,700 moral dilemmas designed to investigate LLMs' decision-making in prioritizing various levels of human needs. In our setup, LLMs act as immigration inspectors deciding whether to approve or deny entry based on the short narratives of people. These narratives are constructed using the Existence, Relatedness, and Growth (ERG) theory, which categorizes human needs into three hierarchical levels. Our analysis of six LLMs reveals statistically significant patterns in decision-making, suggesting that LLMs encode implicit preferences. Additionally, our evaluation of the impact of incorporating social identities into the narratives shows varying responsiveness based on both motivational needs and identity cues, with some models exhibiting higher denial rates for marginalized identities. All data is publicly available at https://github.com/yeonsuuuu28/papers-please.
Authors: Weimin Xiong, Ke Wang, Yifan Song, Hanchao Liu, Sai Zhou, Wei Peng, Sujian Li
Abstract: Current evaluations of tool-integrated LLM agents typically focus on end-to-end tool-usage evaluation while neglecting their stability. This limits their real-world applicability, as various internal or external factors can cause agents to crash or behave abnormally. Our research addresses this by investigating whether agents are vulnerable to errors throughout the entire tool invocation process, including reading tool documentation, selecting tools and generating parameters, and processing the tool's response. Through extensive experiments, we observe that agents are highly susceptible to errors at each stage and agents based on open-source models are more vulnerable than those based on proprietary models. We also find that increasing the model size does not significantly improve tool invocation reasoning and may make agents more vulnerable to attacks resembling normal user instructions. This highlights the importance of evaluating agent stability and offers valuable insights for future LLM development and evaluation.
Authors: Mohamed Ahmed, Mohamed Abdelmouty, Mingyu Kim, Gunvanth Kandula, Alex Park, James C. Davis
Abstract: The advancement of Pre-Trained Language Models (PTLMs) and Large Language Models (LLMs) has led to their widespread adoption across diverse applications. Despite their success, these models remain vulnerable to attacks that exploit their inherent weaknesses to bypass safety measures. Two primary inference-phase threats are token-level and prompt-level jailbreaks. Token-level attacks embed adversarial sequences that transfer well to black-box models like GPT but leave detectable patterns and rely on gradient-based token optimization, whereas prompt-level attacks use semantically structured inputs to elicit harmful responses yet depend on iterative feedback that can be unreliable. To address the complementary limitations of these methods, we propose two hybrid approaches that integrate token- and prompt-level techniques to enhance jailbreak effectiveness across diverse PTLMs. GCG + PAIR and the newly explored GCG + WordGame hybrids were evaluated across multiple Vicuna and Llama models. GCG + PAIR consistently raised attack-success rates over its constituent techniques on undefended models; for instance, on Llama-3, its Attack Success Rate (ASR) reached 91.6%, a substantial increase from PAIR's 58.4% baseline. Meanwhile, GCG + WordGame matched the raw performance of WordGame maintaining a high ASR of over 80% even under stricter evaluators like Mistral-Sorry-Bench. Crucially, both hybrids retained transferability and reliably pierced advanced defenses such as Gradient Cuff and JBShield, which fully blocked single-mode attacks. These findings expose previously unreported vulnerabilities in current safety stacks, highlight trade-offs between raw success and defensive robustness, and underscore the need for holistic safeguards against adaptive adversaries.
Authors: Simon M\"unker, Nils Schwager, Achim Rettinger
Abstract: The ability of Large Language Models (LLMs) to mimic human behavior triggered a plethora of computational social science research, assuming that empirical studies of humans can be conducted with AI agents instead. Since there have been conflicting research findings on whether and when this hypothesis holds, there is a need to better understand the differences in their experimental designs. We focus on replicating the behavior of social network users with the use of LLMs for the analysis of communication on social networks. First, we provide a formal framework for the simulation of social networks, before focusing on the sub-task of imitating user communication. We empirically test different approaches to imitate user behavior on X in English and German. Our findings suggest that social simulations should be validated by their empirical realism measured in the setting in which the simulation components were fitted. With this paper, we argue for more rigor when applying generative-agent-based modeling for social simulation.
Authors: Kartheek Kumar Reddy Nareddy, Sarah Ternus, Julia Niebling
Abstract: The developments in transformer encoder-decoder architectures have led to significant breakthroughs in machine translation, Automatic Speech Recognition (ASR), and instruction-based chat machines, among other applications. The pre-trained models were trained on vast amounts of generic data over a few epochs (fewer than five in most cases), resulting in their strong generalization capabilities. Nevertheless, the performance of these models does suffer when applied to niche domains like transcribing pilot speech in the cockpit, which involves a lot of specific vocabulary and multilingual conversations. This paper investigates and improves the transcription accuracy of cockpit conversations with Whisper models. We have collected around 85 minutes of cockpit simulator recordings and 130 minutes of interview recordings with pilots and manually labeled them. The speakers are middle aged men speaking both German and English. To improve the accuracy of transcriptions, we propose multiple normalization schemes to refine the transcripts and improve Word Error Rate (WER). We then employ fine-tuning to enhance ASR performance, utilizing performance-efficient fine-tuning with Low-Rank Adaptation (LoRA). Hereby, WER decreased from 68.49 \% (pretrained whisper Large model without normalization baseline) to 26.26\% (finetuned whisper Large model with the proposed normalization scheme).
Authors: Delu Kong, Lieve Macken
Abstract: This study focuses on evaluating the performance of machine translations (MTs) compared to human translations (HTs) in English-to-Chinese children's literature translation (CLT) from a stylometric perspective. The research constructs a Peter Pan corpus, comprising 21 translations: 7 human translations (HTs), 7 large language model translations (LLMs), and 7 neural machine translation outputs (NMTs). The analysis employs a generic feature set (including lexical, syntactic, readability, and n-gram features) and a creative text translation (CTT-specific) feature set, which captures repetition, rhythm, translatability, and miscellaneous levels, yielding 447 linguistic features in total. Using classification and clustering techniques in machine learning, we conduct a stylometric analysis of these translations. Results reveal that in generic features, HTs and MTs exhibit significant differences in conjunction word distributions and the ratio of 1-word-gram-YiYang, while NMTs and LLMs show significant variation in descriptive words usage and adverb ratios. Regarding CTT-specific features, LLMs outperform NMTs in distribution, aligning more closely with HTs in stylistic characteristics, demonstrating the potential of LLMs in CLT.
Authors: Delu Kong, Lieve Macken
Abstract: This study explores Machine Translationese (MTese) -- the linguistic peculiarities of machine translation outputs -- focusing on the under-researched English-to-Chinese language pair in news texts. We construct a large dataset consisting of 4 sub-corpora and employ a comprehensive five-layer feature set. Then, a chi-square ranking algorithm is applied for feature selection in both classification and clustering tasks. Our findings confirm the presence of MTese in both Neural Machine Translation systems (NMTs) and Large Language Models (LLMs). Original Chinese texts are nearly perfectly distinguishable from both LLM and NMT outputs. Notable linguistic patterns in MT outputs are shorter sentence lengths and increased use of adversative conjunctions. Comparing LLMs and NMTs, we achieve approximately 70% classification accuracy, with LLMs exhibiting greater lexical diversity and NMTs using more brackets. Additionally, translation-specific LLMs show lower lexical diversity but higher usage of causal conjunctions compared to generic LLMs. Lastly, we find no significant differences between LLMs developed by Chinese firms and their foreign counterparts.
Authors: Baohao Liao, Xinyi Chen, Sara Rajaee, Yuhui Xu, Christian Herold, Anders S{\o}gaard, Maarten de Rijke, Christof Monz
Abstract: Recent advancements in large language models (LLMs) have significantly advanced complex reasoning capabilities, particularly through extended chain-of-thought (CoT) reasoning that incorporates mechanisms such as backtracking, self-reflection and self-correction. Despite these developments, the self-correction abilities of LLMs during long CoT reasoning remain underexplored. And recent findings on overthinking suggest that such models often engage in unnecessarily redundant reasoning. In this work, we empirically show that the first reasoning step exerts a disproportionately large influence on the final prediction - errors introduced at this stage can substantially degrade subsequent reasoning quality. This phenomenon is consistently observed across two state-of-the-art open-source reasoning model families: DeepSeek-R1 and Qwen3. To address this, we propose an efficient sampling strategy that leverages a reward model to identify and retain high-quality first reasoning steps while discarding suboptimal ones, achieving up to a 70% reduction in inference cost without sacrificing accuracy. Finally, we introduce a new benchmark specifically constructed with deliberately flawed first reasoning steps to systematically evaluate model self-correction capabilities, offering a foundation for future research on robust reasoning in LLMs.
Authors: Chris Madge, Maris Camilleri, Paloma Carretero Garcia, Mladen 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: Eleonora Amadori, Daniele Cirulli, Edoardo Di Martino, Jacopo Nudo, Maria Sahakyan, Emanuele Sangiorgio, Arnaldo Santoro, Simon Zollo, Alessandro Galeazzi, Niccol\`o Di Marco
Abstract: Language is a fundamental aspect of human societies, continuously evolving in response to various stimuli, including societal changes and intercultural interactions. Technological advancements have profoundly transformed communication, with social media emerging as a pivotal force that merges entertainment-driven content with complex social dynamics. As these platforms reshape public discourse, analyzing the linguistic features of user-generated content is essential to understanding their broader societal impact. In this paper, we examine the linguistic complexity of content produced by influential users on Twitter across three globally significant and contested topics: COVID-19, COP26, and the Russia-Ukraine war. By combining multiple measures of textual complexity, we assess how language use varies along four key dimensions: account type, political leaning, content reliability, and sentiment. Our analysis reveals significant differences across all four axes, including variations in language complexity between individuals and organizations, between profiles with sided versus moderate political views, and between those associated with higher versus lower reliability scores. Additionally, profiles producing more negative and offensive content tend to use more complex language, with users sharing similar political stances and reliability levels converging toward a common jargon. Our findings offer new insights into the sociolinguistic dynamics of digital platforms and contribute to a deeper understanding of how language reflects ideological and social structures in online spaces.
Authors: David Demitri Africa
Abstract: I implement a procedure to isolate and interpret the sub-network (or "circuit") responsible for subject-verb agreement in GPT-2 Small. In this study, the model is given prompts where the subject is either singular (e.g. "Alice") or plural (e.g. "Alice and Bob"), and the task is to correctly predict the appropriate verb form ("walks" for singular subjects, "walk" for plural subjects). Using a series of techniques-including performance verification automatic circuit discovery via direct path patching, and direct logit attribution- I isolate a candidate circuit that contributes significantly to the model's correct verb conjugation. The results suggest that only a small fraction of the network's component-token pairs is needed to achieve near-model performance on the base task but substantially more for more complex settings.
Authors: Iliass Ayaou (ICube), Denis Cavallucci (ICube), Hicham Chibane (ICube)
Abstract: In the landscape of publicly available patent retrieval datasets, the need for explicit indomain and out-of-domain labeling, multi-jurisdiction coverage, balanced query domain representation and manageable sizes that support sub document level experiments on moderate computational resources is often overlooked. To address these gaps, we propose DAPFAM, a new open access domain-aware patent retrieval dataset constructed at the simple-family level. The dataset contains 1,247 domain balanced full text query families and 45,336 full text target families. The dataset is enriched by clear relevance judgments (forward/backward citations as positive links, random negatives), as well as explicit in-domain or out-of-domain relationships via a novel proposed labelling scheme based on via International Patent Classification (IPC) codes, resulting in 49,869 evaluation pairs. The dataset is multi jurisdictional, requires little to no preprocessing for retrieval evaluation, and remains of a size manageable for entities with limited ressources allowing for sub document level retrieval experiments without excessive computational costs. We describe our three-step data-curation pipeline, present comprehensive dataset statistics, and provide baseline experiments using lexical and neural retrieval methods. Our baseline experiments highlight significant challenges in crossdomain patent retrieval. The dataset will be publicly available (for now the access link is this repository: https://osf.io/vbyzd/?view_only=1a40242e0d1941a58aa854af3e50cf6b).
URLs: https://osf.io/vbyzd/?view_only=1a40242e0d1941a58aa854af3e50cf6b).
Authors: Muhammad Umar Farooq, Oscar Saz
Abstract: This paper investigates the performance of various speech SSL models on dialectal Arabic (DA) and Arabic-English code-switched (CS) speech. To address data scarcity, a modified audio-splicing approach is introduced to generate artificial CS speech data. Fine-tuning an already fine-tuned SSL model with the proposed Spliced-Audio Generated (SAGE) data results in an absolute improvement on Word Error Rate (WER) of 7.8% on Arabic and English CS benchmarks. Additionally, an Experience Replay (ER) inspired approach is proposed to enhance generalisation across DA and CS speech while mitigating catastrophic forgetting. Integrating an out-of-domain 3-gram language model reduces the overall mean WER from 31.7% to 26.6%. Few-shot fine-tuning for code-switching benchmarks further improves WER by 4.9%. A WER of 31.1% on Arabic-English CS benchmarks surpasses large-scale multilingual models, including USM and Whisper-large-v2 (both over ten times larger) by an absolute margin of 5.5% and 8.4%, respectively.
Authors: Tianshu Yu, Chao Xiang, Mingchuan Yang, Pei Ke, Bosi Wen, Cunxiang Wang, Jiale Cheng, Li Zhang, Xinyu Mu, Chuxiong Sun, Minlie Huang
Abstract: Large language models (LLMs) have demonstrated remarkable evaluation and critique capabilities, providing insightful feedback and identifying flaws in various tasks. However, limited research has explored which types of critiques are most effective for improving model responses or how to generate such critiques. To address this gap, we introduce \textbf{R}efinement-oriented \textbf{C}ritique \textbf{O}ptimization (RCO), a novel framework designed to train critic models using refinement signals. RCO uses a feedback loop where critiques, generated by the critic model, guide the actor model in refining its responses. The critique utility (CU) quantifies the effectiveness of these refinements, serving as the reward signal for training the critic model. By focusing on critiques that lead to better refinements, RCO eliminates the need for direct critique preference assessment, ensuring that critiques driving meaningful improvements are rewarded. We evaluate RCO across five tasks, i.e., dialog generation, summarization, question answering, mathematical reasoning, and code generation, and show that it significantly outperforms traditional methods and open-source models in terms of critique quality and refinement outcomes. Our contributions include the introduction of RCO, a novel supervision scheme based on refined response preferences, and comprehensive experimental results that highlight the method's effectiveness in enhancing LLM critique-refinement loops.
Authors: Patrick Haller, Jannis Vamvas, Rico Sennrich, Lena A. J\"ager
Abstract: A growing body of work has been querying LLMs with political questions to evaluate their potential biases. However, this probing method has limited stability, making comparisons between models unreliable. In this paper, we argue that LLMs need more context. We propose a new probing task, Questionnaire Modeling (QM), that uses human survey data as in-context examples. We show that QM improves the stability of question-based bias evaluation, and demonstrate that it may be used to compare instruction-tuned models to their base versions. Experiments with LLMs of various sizes indicate that instruction tuning can indeed change the direction of bias. Furthermore, we observe a trend that larger models are able to leverage in-context examples more effectively, and generally exhibit smaller bias scores in QM. Data and code are publicly available.
Authors: Albert Agisha Ntwali, Luca R\"uck, Martin Heckmann
Abstract: We propose a novel approach for detecting personal data in structured datasets, leveraging GPT-4o, a state-of-the-art Large Language Model. A key innovation of our method is the incorporation of contextual information: in addition to a feature's name and values, we utilize information from other feature names within the dataset as well as the dataset description. We compare our approach to alternative methods, including Microsoft Presidio and CASSED, evaluating them on multiple datasets: DeSSI, a large synthetic dataset, datasets we collected from Kaggle and OpenML as well as MIMIC-Demo-Ext, a real-world dataset containing patient information from critical care units. Our findings reveal that detection performance varies significantly depending on the dataset used for evaluation. CASSED excels on DeSSI, the dataset on which it was trained. Performance on the medical dataset MIMIC-Demo-Ext is comparable across all models, with our GPT-4o-based approach clearly outperforming the others. Notably, personal data detection in the Kaggle and OpenML datasets appears to benefit from contextual information. This is evidenced by the poor performance of CASSED and Presidio (both of which do not utilize the context of the dataset) compared to the strong results of our GPT-4o-based approach. We conclude that further progress in this field would greatly benefit from the availability of more real-world datasets containing personal information.
Authors: Qingquan Li, Shaoyu Dou, Kailai Shao, Chao Chen, Haixiang Hu
Abstract: The remarkable performance of Large Language Models (LLMs) gives rise to``LLM-as-a-Judge'', where LLMs are employed as evaluators for complex tasks. Moreover, it has been widely adopted across fields such as Natural Language Processing (NLP), preference learning, and various specific domains. However, there are various biases within LLM-as-a-Judge, which adversely affect the fairness and reliability of judgments. Current research on evaluating or mitigating bias in LLM-as-a-Judge predominantly focuses on comparison-based evaluations, while systematic investigations into bias in scoring-based evaluations remain limited. Therefore, we define scoring bias in LLM-as-a-Judge as the scores differ when scoring judge models are bias-related perturbed, and provide a well-designed framework to comprehensively evaluate scoring bias. We augment existing LLM-as-a-Judge benchmarks through data synthesis to construct our evaluation dataset and design multi-faceted evaluation metrics. Our experimental results demonstrate that the scoring stability of existing judge models is disrupted by scoring biases. Further exploratory experiments and discussions provide valuable insights into the design of scoring prompt templates and the mitigation of scoring biases on aspects such as score rubrics, score IDs, and reference answer selection.
Authors: Daichi Kato, Ryo Ueda, Yusuke Miyao
Abstract: If humans understood language by randomly selecting parsing actions, it might have been necessary to construct a robust symbolic system capable of being interpreted under any hierarchical structure. However, human parsing strategies do not seem to follow such a random pattern. Why is that the case? In fact, a previous study on emergent communication using models with hierarchical biases have reported that agents adopting random parsing strategies$\unicode{x2013}$ones that deviate significantly from human language comprehension$\unicode{x2013}$can achieve high communication accuracy. In this study, we investigate this issue by making two simple and natural modifications to the experimental setup: (I) we use more complex inputs that have hierarchical structures, such that random parsing makes semantic interpretation more difficult, and (II) we incorporate a surprisal-related term, which is known to influence the order of words and characters in natural language, into the objective function. With these changes, we evaluate whether agents employing random parsing strategies still maintain high communication accuracy.
Authors: Danush Khanna, Aditya Kumar Guru, Srivarshinee Sridhar, Zidan Ahmed, Rubhav Bahirwani, Meetu Malhotra, Vinija Jain, Aman Chadha, Amitava Das, Kripabandhu Ghosh
Abstract: Inference accounts for the majority of latency and energy consumption in large language model (LLM) deployments, often exceeding 90% of total cost. While training-time efficiency has seen extensive progress, runtime optimization remains a key bottleneck, particularly under autoregressive decoding. Existing approaches -- such as pruning, quantization, early exits, and speculative decoding -- often require retraining, architectural changes, or disrupt decoding compatibility. We introduce QuickSilver, a modular, token-level framework that enables semantic adaptivity at inference time without altering model weights or structure. QuickSilver integrates four synergistic mechanisms: (i) Dynamic Token Halting, which halts computation for tokens with converged representations; (ii) KV Cache Skipping, which selectively suppresses memory writes to reduce attention overhead; and (iii) Contextual Token Fusion, which collapses redundant tokens into shared paths to shrink sequence length. Unlike speculative decoding or MoE routing, QuickSilver operates entirely on frozen, dense models and requires no auxiliary networks. Applied to GPT-2 and Llama-2 across WikiText-103 and C4, QuickSilver achieves up to 39.6% FLOP reduction with negligible perplexity degradation (<=0.2).
Authors: Petr Pechman, Milan Straka, Jana Strakov\'a, Jakub N\'aplava
Abstract: We present a grammar error correction (GEC) system that achieves state of the art for the Czech language. Our system is based on a neural network translation approach with the Transformer architecture, and its key feature is its real-time synthetic generation pipeline, which dynamically augments sentences with artificial errors by introducing both language-agnostic and Czech-specific errors. We conduct a comprehensive series of experiments, investigating the Czech GEC corpora as bases for synthetic error introduction, several error generation strategies, domain balancing, tokenization granularity, model size, and data scaling during fine-tuning. Additionally, we evaluate the performance of large language models (LLMs) on Czech GEC in both end-user and expert fine-tuning scenarios. Our best-performing model is superior both in performance and computational efficiency. The source code and the trained model links are available on https://github.com/ufal/tsd2025-gec.
Authors: NAVER Cloud HyperCLOVA X Team
Abstract: We introduce HyperCLOVA X THINK, the first reasoning-focused large language model in the HyperCLOVA X family, pre-trained on roughly $6$ trillion high-quality Korean, and English tokens, augmented with targeted synthetic Korean data. It was implemented as a compute-memory-balanced Peri-LN Transformer scaled with $\mu$P, pre-trained through a three-stage curriculum that expands the context window to $128$K tokens, and post-trained via supervised fine-tuning with Reinforcement Learning from Verifiable Rewards supports both detailed rationale and concise-answer modes. It delivers competitive performance against similarly sized models on Korea-focused benchmarks such as KMMLU, CSAT, KoBALT-700, HAERAE-1.0, and KoBigBench, while preserving robust bilingual consistency and translation quality. In addition, a vision-augmented variant matches or exceeds GPT-4.1 on the KCSAT STEM benchmark, all of which are achieved with substantially lower training compute than existing models of similar sizes. We also present a pruning and distillation technique that will soon be applied to HyperCLOVA X THINK for an open-source and business-friendly foundation model. Altogether, these capabilities position HyperCLOVA X THINK as a robust foundation for Korean AI innovation and a valuable resource for the global research community.
Authors: Harsha Nori, Mayank Daswani, Christopher Kelly, Scott Lundberg, Marco Tulio Ribeiro, Marc Wilson, Xiaoxuan Liu, Viknesh Sounderajah, Jonathan Carlson, Matthew P Lungren, Bay Gross, Peter Hames, Mustafa Suleyman, Dominic King, Eric Horvitz
Abstract: Artificial intelligence holds great promise for expanding access to expert medical knowledge and reasoning. However, most evaluations of language models rely on static vignettes and multiple-choice questions that fail to reflect the complexity and nuance of evidence-based medicine in real-world settings. In clinical practice, physicians iteratively formulate and revise diagnostic hypotheses, adapting each subsequent question and test to what they've just learned, and weigh the evolving evidence before committing to a final diagnosis. To emulate this iterative process, we introduce the Sequential Diagnosis Benchmark, which transforms 304 diagnostically challenging New England Journal of Medicine clinicopathological conference (NEJM-CPC) cases into stepwise diagnostic encounters. A physician or AI begins with a short case abstract and must iteratively request additional details from a gatekeeper model that reveals findings only when explicitly queried. Performance is assessed not just by diagnostic accuracy but also by the cost of physician visits and tests performed. We also present the MAI Diagnostic Orchestrator (MAI-DxO), a model-agnostic orchestrator that simulates a panel of physicians, proposes likely differential diagnoses and strategically selects high-value, cost-effective tests. When paired with OpenAI's o3 model, MAI-DxO achieves 80% diagnostic accuracy--four times higher than the 20% average of generalist physicians. MAI-DxO also reduces diagnostic costs by 20% compared to physicians, and 70% compared to off-the-shelf o3. When configured for maximum accuracy, MAI-DxO achieves 85.5% accuracy. These performance gains with MAI-DxO generalize across models from the OpenAI, Gemini, Claude, Grok, DeepSeek, and Llama families. We highlight how AI systems, when guided to think iteratively and act judiciously, can advance diagnostic precision and cost-effectiveness in clinical care.
Authors: Yifan Shen, Yuanzhe Liu, Jingyuan Zhu, Xu Cao, Xiaofeng Zhang, Yixiao He, Wenming Ye, James Matthew Rehg, Ismini Lourentzou
Abstract: Current Vision-Language Models (VLMs) struggle with fine-grained spatial reasoning, particularly when multi-step logic and precise spatial alignment are required. In this work, we introduce SpatialReasoner-R1, a vision-language reasoning model designed to address these limitations. To construct high-quality supervision for spatial reasoning, we design a Multi-Model Monte Carlo Tree Search (M3CTS) method that generates diverse, logically consistent Long Chain-of-Thought (LongCoT) reasoning trajectories. In addition, we propose fine-grained Direct Preference Optimization (fDPO), which introduces segment-specific preference granularity for descriptive grounding and logical reasoning, guided by a spatial reward mechanism that evaluates candidate responses based on visual consistency, spatial grounding, and logical coherence. Experimental results demonstrate that fDPO achieves an average improvement of 4.1% over standard DPO across spatial quality tasks, and a 9.0% gain in spatial quantity tasks. SpatialReasoner-R1, trained with fDPO, sets a new SoTA on SPATIALRGPT-Bench, outperforming the strongest baseline by 9.8% in average accuracy, while maintaining competitive performance on general vision-language tasks.
Authors: Nicolas Bougie, Narimasa Watanabe
Abstract: Modeling human behavior in urban environments is fundamental for social science, behavioral studies, and urban planning. Prior work often rely on rigid, hand-crafted rules, limiting their ability to simulate nuanced intentions, plans, and adaptive behaviors. Addressing these challenges, we envision an urban simulator (CitySim), capitalizing on breakthroughs in human-level intelligence exhibited by large language models. In CitySim, agents generate realistic daily schedules using a recursive value-driven approach that balances mandatory activities, personal habits, and situational factors. To enable long-term, lifelike simulations, we endow agents with beliefs, long-term goals, and spatial memory for navigation. CitySim exhibits closer alignment with real humans than prior work, both at micro and macro levels. Additionally, we conduct insightful experiments by modeling tens of thousands of agents and evaluating their collective behaviors under various real-world scenarios, including estimating crowd density, predicting place popularity, and assessing well-being. Our results highlight CitySim as a scalable, flexible testbed for understanding and forecasting urban phenomena.
Authors: Abdulkareem Alsudais
Abstract: The rise and growing popularity of accessible large language models have raised questions about their impact on various aspects of life, including how scientists write and publish their research. The primary objective of this paper is to analyze a dataset consisting of all abstracts posted on arXiv.org between 2010 and June 7th, 2024, to assess the evolution of their readability and determine whether significant shifts occurred following the release of ChatGPT in November 2022. Four standard readability formulas are used to calculate individual readability scores for each paper, classifying their level of readability. These scores are then aggregated by year and across the eight primary categories covered by the platform. The results show a steady annual decrease in readability, suggesting that abstracts are likely becoming increasingly complex. Additionally, following the release of ChatGPT, a significant change in readability is observed for 2023 and the analyzed months of 2024. Similar trends are found across categories, with most experiencing a notable change in readability during 2023 and 2024. These findings offer insights into the broader changes in readability and point to the likely influence of AI on scientific writing.
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: Zhuodi Cai
Abstract: This paper presents 3Description, an experimental human-AI collaborative approach for intuitive 3D modeling. 3Description aims to address accessibility and usability challenges in traditional 3D modeling by enabling non-professional individuals to co-create 3D models using verbal and gesture descriptions. Through a combination of qualitative research, product analysis, and user testing, 3Description integrates AI technologies such as Natural Language Processing and Computer Vision, powered by OpenAI and MediaPipe. Recognizing the web has wide cross-platform capabilities, 3Description is web-based, allowing users to describe the desired model and subsequently adjust its components using verbal and gestural inputs. In the era of AI and emerging media, 3Description not only contributes to a more inclusive and user-friendly design process, empowering more people to participate in the construction of the future 3D world, but also strives to increase human engagement in co-creation with AI, thereby avoiding undue surrender to technology and preserving human creativity.
Authors: Haofeng Wang, Yilin Guo, Zehao Li, Tong Yue, Yizong Wang, Enci Zhang, Rongqun Lin, Feng Gao, Shiqi Wang, Siwei Ma
Abstract: The Yellow River is China's mother river and a cradle of human civilization. The ancient Yellow River culture is, moreover, an indispensable part of human art history. To conserve and inherit the ancient Yellow River culture, we designed RiverEcho, a real-time interactive system that responds to voice queries using a large language model and a cultural knowledge dataset, delivering explanations through a talking-head digital human. Specifically, we built a knowledge database focused on the ancient Yellow River culture, including the collection of historical texts and the processing pipeline. Experimental results demonstrate that leveraging Retrieval-Augmented Generation (RAG) on the proposed dataset enhances the response quality of the Large Language Model(LLM), enabling the system to generate more professional and informative responses. Our work not only diversifies the means of promoting Yellow River culture but also provides users with deeper cultural insights.
Authors: Zunran Wang, Zheng Shenpeng, Wang Shenglan, Minghui Zhao, Zhonghua Li
Abstract: Hybrid-based retrieval methods, which unify dense-vector and lexicon-based retrieval, have garnered considerable attention in the industry due to performance enhancement. However, despite their promising results, the application of these hybrid paradigms in Chinese retrieval contexts has remained largely underexplored. In this paper, we introduce HyReC, an innovative end-to-end optimization method tailored specifically for hybrid-based retrieval in Chinese. HyReC enhances performance by integrating the semantic union of terms into the representation model. Additionally, it features the Global-Local-Aware Encoder (GLAE) to promote consistent semantic sharing between lexicon-based and dense retrieval while minimizing the interference between them. To further refine alignment, we incorporate a Normalization Module (NM) that fosters mutual benefits between the retrieval approaches. Finally, we evaluate HyReC on the C-MTEB retrieval benchmark to demonstrate its effectiveness.
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: Michael A. Riegler, Kristoffer Herland Hellton, Vajira Thambawita, Hugo L. Hammer
Abstract: Selecting prior distributions in Bayesian statistics is challenging, resource-intensive, and subjective. We analyze using large-language models (LLMs) to suggest suitable, knowledge-based informative priors. We developed an extensive prompt asking LLMs not only to suggest priors but also to verify and reflect on their choices. We evaluated Claude Opus, Gemini 2.5 Pro, and ChatGPT-4o-mini on two real datasets: heart disease risk and concrete strength. All LLMs correctly identified the direction for all associations (e.g., that heart disease risk is higher for males). The quality of suggested priors was measured by their Kullback-Leibler divergence from the maximum likelihood estimator's distribution. The LLMs suggested both moderately and weakly informative priors. The moderate priors were often overconfident, resulting in distributions misaligned with the data. In our experiments, Claude and Gemini provided better priors than ChatGPT. For weakly informative priors, a key performance difference emerged: ChatGPT and Gemini defaulted to an "unnecessarily vague" mean of 0, while Claude did not, demonstrating a significant advantage. The ability of LLMs to identify correct associations shows their great potential as an efficient, objective method for developing informative priors. However, the primary challenge remains in calibrating the width of these priors to avoid over- and under-confidence.
Authors: Bohan Li, Zhihan Li, Haoran Wang, Hanglei Zhang, Yiwei Guo, Hankun Wang, Xie Chen, Kai Yu
Abstract: Recently, autoregressive (AR) language models have emerged as a dominant approach in speech synthesis, offering expressive generation and scalable training. However, conventional AR speech synthesis models relying on the next-token prediction paradigm often encounter significant challenges when handling long speech sequences. These models often struggle to construct stable frame-to-frame attention, leading to increased latency and degraded synthesis quality, thereby limiting their feasibility for real-time applications. To address these limitations, we introduce a novel dynamic chunk-wise autoregressive synthesis framework, termed DCAR, designed to enhance both efficiency and intelligibility robustness in AR speech generation. DCAR introduces a chunk-to-frame attention mechanism through training with multi-token prediction, enabling dynamic chunk prediction in variable speech contexts using a lightweight module trained on-policy. DCAR dynamically adjusts the token prediction span, significantly reducing the sequence length dependency while obtaining high synthesis quality. Comprehensive empirical evaluations demonstrate that DCAR substantially outperforms traditional next-token prediction models, achieving up to 72.27% intelligibility improvement and 2.61x inference speedup simultaneously on the test set. Furthermore, we conduct comprehensive analysis to support it as a versatile foundation for next-generation speech synthesis systems.
Authors: Tianhao Chen, Xin Xu, Zijing Liu, Pengxiang Li, Xinyuan Song, Ajay Kumar Jaiswal, Fan Zhang, Jishan Hu, Yang Wang, Hao Chen, Shizhe Diao, Shiwei Liu, Yu Li, Yin Lu, Can Yang
Abstract: Modern Large Language Models, such as the LLaMA, Qwen and DeepSeek series, predominantly adopt the Pre-LayerNorm (Pre-LN) Transformer architecture. While being stable during pretraining and scalable to large model sizes, Pre-LN suffers from an exponential growth in activation variance across layers, causing the residual path to dominate over sub-layer outputs and limiting the learning capacity of deeper layers. To mitigate this issue, we propose Gradient-Preserving Activation Scaling (GPAS), a simple technique that can be used in combination with existing approaches. GPAS works by scaling down the intermediate activations while keeping their gradients unchanged. This leaves information in the activations intact, and avoids the gradient vanishing problem associated with gradient downscaling. Extensive experiments across various model sizes from 71M to 1B show that GPAS achieves consistent performance gains. Beyond enhancing Pre-LN Transformers, GPAS also shows promise in improving alternative architectures such as Sandwich-LN and DeepNorm, demonstrating its versatility and potential for improving training dynamics in a wide range of settings.
Authors: Laura van Weesep, Samuel Genheden, Ola Engkvist, Jens Sj\"olund
Abstract: Large-language models (LLMs) and agentic systems present exciting opportunities to accelerate drug discovery and design. In this study, we critically examine the modularity of LLM-based agentic systems for drug discovery, i.e., whether parts of the agentic system such as the LLM are interchangeable, a topic that has received limited attention in drug discovery applications. We compare the performance of different large language models (LLMs) and the effectiveness of tool-calling agents versus code-generating agents in this domain. Our case study, comparing performance in orchestrating tools for chemistry and drug discovery using an LLM-as-a-judge score, shows that Claude-3.5-Sonnet, Claude-3.7-Sonnet and GPT-4o outperform alternative language models such as Llama-3.1-8B, Llama-3.1-70B, GPT-3.5-Turbo, and Nova-Micro. Although we confirm that code-generating agents outperform the tool-calling ones on average, we show that this is highly question and model dependent. Furthermore, the impact of replacing system prompts is dependent on the specific question asked and the model used, underscoring that -- even in this particular domain -- one cannot just replace language models without considering prompt re-engineering. Our study highlights the necessity of further research into the modularity of agentic systems to enable the development of stable and scalable solutions for real-world problems.
Authors: Sebastian Murgul, Moritz Reiser, Michael Heizmann, Christoph Seibert
Abstract: In this paper, we present a neural network approach for synchronizing audio recordings of human piano performances with their corresponding loosely aligned MIDI files. The task is addressed using a Convolutional Recurrent Neural Network (CRNN) architecture, which effectively captures spectral and temporal features by processing an unaligned piano roll and a spectrogram as inputs to estimate the aligned piano roll. To train the network, we create a dataset of piano pieces with augmented MIDI files that simulate common human timing errors. The proposed model achieves up to 20% higher alignment accuracy than the industry-standard Dynamic Time Warping (DTW) method across various tolerance windows. Furthermore, integrating DTW with the CRNN yields additional improvements, offering enhanced robustness and consistency. These findings demonstrate the potential of neural networks in advancing state-of-the-art MIDI-to-audio alignment.
Authors: Maciej Stefaniak, Micha{\l} Krutul, Jan Ma{\l}a\'snicki, Maciej Pi\'oro, Jakub Krajewski, Sebastian Jaszczur, Marek Cygan, Kamil Adamczewski, Jan Ludziejewski
Abstract: Large language models have steadily increased in size to achieve improved performance; however, this growth has also led to greater inference time and computational demands. Consequently, there is rising interest in model size reduction methods. To address this issue, we propose Projected Compression, a novel model compression technique, that reduces model weights by utilizing projection modules. Specifically, we first train additional trainable projections weights and preserve access to all the original model parameters. Subsequently, these projections are merged into a lower-dimensional product matrix, resulting in a reduced-size standard Transformer-based model. Unlike alternative approaches that require additional computational overhead, our method matches the base model's per-token computation step in FLOPs. Experimental results show that Projected Compression outperforms the comparable hard pruning and retraining approach on higher quality models. Moreover, the performance margin scales well with the number of tokens.
Authors: Filippo Merlo, Ece Takmaz, Wenkai Chen, Albert Gatt
Abstract: Natural scenes provide us with rich contexts for object recognition and reference. In particular, knowing what type of scene one is looking at generates expectations about which objects will occur, and what their spatial configuration should be. Do Vision-Language Models (VLMs) learn to rely on scene contexts in a similar way, when generating references to objects? To address this question, we introduce the \textit{Common Objects Out-of-Context (COOCO)} dataset and test to what extent VLMs rely on scene context to refer to objects under different degrees of scene-object congruency, and different perturbations. Our findings show that models leverage scene context adaptively, depending on both the semantic relatedness between object and scene and the level of noise. In particular, models rely more on context under high target-scene congruence or when objects are degraded. Attention analysis reveals that successful object categorisation involves increased focus on the target in mid-level layers, especially under moderate noise, suggesting that VLMs dynamically balance local and contextual information for reference generation. We make our dataset, code and models available at \href{https://github.com/cs-nlp-uu/scenereg}{https://github.com/cs-nlp-uu/scenereg}.
URLs: https://github.com/cs-nlp-uu/scenereg, https://github.com/cs-nlp-uu/scenereg
Authors: Klara M. Gutekunst, Dominik D\"urrschnabel, Johannes Hirth, Gerd Stumme
Abstract: The vast growth of data has rendered traditional manual inspection infeasible, necessitating the adoption of computational methods for efficient data exploration. Topic modeling has emerged as a powerful tool for analyzing large-scale textual datasets, enabling the extraction of latent semantic structures. However, existing methods for topic modeling often struggle to provide interpretable representations that facilitate deeper insights into data structure and content. In this paper, we propose FAT-CAT, an approach based on Formal Concept Analysis (FCA) to enhance meaningful topic aggregation and visualization of discovered topics. Our approach can handle diverse topics and file types -- grouped by directories -- to construct a concept lattice that offers a structured, hierarchical representation of their topic distribution. In a case study on the ETYNTKE dataset, we evaluate the effectiveness of our approach against other representation methods to demonstrate that FCA-based aggregation provides more meaningful and interpretable insights into dataset composition than existing topic modeling techniques.
Authors: Xiang Li, Garrett Wen, Weiqing He, Jiayuan Wu, Qi Long, Weijie J. Su
Abstract: Text watermarks in large language models (LLMs) are an increasingly important tool for detecting synthetic text and distinguishing human-written content from LLM-generated text. While most existing studies focus on determining whether entire texts are watermarked, many real-world scenarios involve mixed-source texts, which blend human-written and watermarked content. In this paper, we address the problem of optimally estimating the watermark proportion in mixed-source texts. We cast this problem as estimating the proportion parameter in a mixture model based on \emph{pivotal statistics}. First, we show that this parameter is not even identifiable in certain watermarking schemes, let alone consistently estimable. In stark contrast, for watermarking methods that employ continuous pivotal statistics for detection, we demonstrate that the proportion parameter is identifiable under mild conditions. We propose efficient estimators for this class of methods, which include several popular unbiased watermarks as examples, and derive minimax lower bounds for any measurable estimator based on pivotal statistics, showing that our estimators achieve these lower bounds. Through evaluations on both synthetic data and mixed-source text generated by open-source models, we demonstrate that our proposed estimators consistently achieve high estimation accuracy.
Authors: Maryam Mousavian, Zahra Abbasiantaeb, Mohammad Aliannejadi, Fabio Crestani
Abstract: The presence of social biases in Natural Language Processing (NLP) and Information Retrieval (IR) systems is an ongoing challenge, which underlines the importance of developing robust approaches to identifying and evaluating such biases. In this paper, we aim to address this issue by leveraging Large Language Models (LLMs) to detect and measure gender bias in passage ranking. Existing gender fairness metrics rely on lexical- and frequency-based measures, leading to various limitations, e.g., missing subtle gender disparities. Building on our LLM-based gender bias detection method, we introduce a novel gender fairness metric, named Class-wise Weighted Exposure (CWEx), aiming to address existing limitations. To measure the effectiveness of our proposed metric and study LLMs' effectiveness in detecting gender bias, we annotate a subset of the MS MARCO Passage Ranking collection and release our new gender bias collection, called MSMGenderBias, to foster future research in this area. Our extensive experimental results on various ranking models show that our proposed metric offers a more detailed evaluation of fairness compared to previous metrics, with improved alignment to human labels (58.77% for Grep-BiasIR, and 18.51% for MSMGenderBias, measured using Cohen's Kappa agreement), effectively distinguishing gender bias in ranking. By integrating LLM-driven bias detection, an improved fairness metric, and gender bias annotations for an established dataset, this work provides a more robust framework for analyzing and mitigating bias in IR systems.
Authors: Youkang Wang, Jian Wang, Rubing Chen, Xiao-Yong Wei, Qing Li
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 \textsc{OptScale}, a practical algorithm that dynamically determines the optimal number of sampled responses. \textsc{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 \textsc{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.
Authors: Yue Zhang, Jilei Sun, Yunhui Guo, Vibhav Gogate
Abstract: Video Large Multimodal Models (VLMMs) have made impressive strides in understanding video content, but they often struggle with abstract and adaptive reasoning-the ability to revise their interpretations when new information emerges. In reality, conclusions are rarely set in stone; additional context can strengthen or weaken an initial inference. To address this, we introduce Defeasible Video Entailment (DVidE), a new task that challenges models to think like doubters, constantly updating their reasoning based on evolving evidence. In DVidE, given a video premise and a textual hypothesis, models must determine whether a new update strengthens or weakens the hypothesis (classification version) or generate a coherent update that modifies the entailment relationship (generation version). For solving the classification task, we propose the Chain of Counterfactual Thought framework, utilizing counterfactual reasoning, ASR-enhanced video content, and rationale refinement to reduce inference bias. For the generation task, we develop a framework that combines ASR output with a Large Language Model (LLM) to produce coherent, contextually relevant updates aligned with the intended strengthener or weakener goals. Additionally, we introduce a novel benchmark dataset, with strengthener/weakener annotations and an LLM-based evaluation metric specifically designed for assessing generative performance. Experimental results demonstrate significant improvements, highlighting our proposed method in enhancing dynamic reasoning capabilities of VLMMs.
Authors: Bingchen Zhao, Despoina Magka, Minqi Jiang, Xian Li, Roberta Raileanu, Tatiana Shavrina, Jean-Christophe Gagnon-Audet, Kelvin Niu, Shagun Sodhani, Michael Shvartsman, Andrei Lupu, Alisia Lupidi, Edan Toledo, Karen Hambardzumyan, Martin Josifoski, Thomas Foster, Lucia Cipolina-Kun, Abhishek Charnalia, Derek Dunfield, Alexander H. Miller, Oisin Mac Aodha, Jakob Foerster, Yoram Bachrach
Abstract: Rapid advancements in large language models (LLMs) have the potential to assist in scientific progress. A critical capability toward this endeavor is the ability to reproduce existing work. To evaluate the ability of AI agents to reproduce results in an active research area, we introduce the Automated LLM Speedrunning Benchmark, leveraging the research community contributions on the NanoGPT speedrun, a competition to train a GPT-2 model in the shortest time. Each of the 19 speedrun tasks provides the agent with the previous records training script, optionally paired with one of three hint formats, ranging from pseudocode to paper-like descriptions of the new records improvements. Records execute quickly by design and speedrun improvements encompass diverse code-level changes, ranging from high-level algorithmic advancements to hardware-aware optimizations. These features make the benchmark both accessible and realistic for the frontier problem of improving LLM training. We find that recent reasoning LLMs combined with SoTA scaffolds struggle to reimplement already-known innovations in our benchmark, even when given detailed hints. Our benchmark thus provides a simple, non-saturated measure of an LLMs ability to automate scientific reproduction, a necessary (but not sufficient) skill for an autonomous research agent.
Authors: James Chua, Edward Rees, Hunar Batra, Samuel R. Bowman, Julian Michael, Ethan Perez, Miles Turpin
Abstract: Chain-of-thought prompting (CoT) has the potential to improve the explainability of language model reasoning. But CoT can also systematically misrepresent the factors influencing models' behavior -- for example, rationalizing answers in line with a user's opinion. We first create a new dataset of 9 different biases that affect GPT-3.5-Turbo and Llama-8b models. These consist of spurious-few-shot patterns, post hoc rationalization, and sycophantic settings. Models switch to the answer implied by the bias, without mentioning the effect of the bias in the CoT. To mitigate this biased reasoning problem, we introduce bias-augmented consistency training (BCT), an unsupervised fine-tuning scheme that trains models to give consistent reasoning across prompts with and without biasing features. We construct a suite testing nine forms of biased reasoning on seven question-answering tasks, and find that applying BCT to GPT-3.5-Turbo with one bias reduces the rate of biased reasoning by 86\% on held-out tasks. Moreover, this model generalizes to other forms of bias, reducing biased reasoning on held-out biases by an average of 37\%. As BCT generalizes to held-out biases and does not require gold labels, this method may hold promise for reducing biased reasoning from as-of-yet unknown biases and on tasks where ground truth reasoning is unavailable.
Authors: Vittoria Dentella, Fritz Guenther, Evelina Leivada
Abstract: Understanding the limits of language is a prerequisite for Large Language Models (LLMs) to act as theories of natural language. LLM performance in some language tasks presents both quantitative and qualitative differences from that of humans, however it remains to be determined whether such differences are amenable to model size. This work investigates the critical role of model scaling, determining whether increases in size make up for such differences between humans and models. We test three LLMs from different families (Bard, 137 billion parameters; ChatGPT-3.5, 175 billion; ChatGPT-4, 1.5 trillion) on a grammaticality judgment task featuring anaphora, center embedding, comparatives, and negative polarity. N=1,200 judgments are collected and scored for accuracy, stability, and improvements in accuracy upon repeated presentation of a prompt. Results of the best performing LLM, ChatGPT-4, are compared to results of n=80 humans on the same stimuli. We find that humans are overall less accurate than ChatGPT-4 (76% vs. 80% accuracy, respectively), but that this is due to ChatGPT-4 outperforming humans only in one task condition, namely on grammatical sentences. Additionally, ChatGPT-4 wavers more than humans in its answers (12.5% vs. 9.6% likelihood of an oscillating answer, respectively). Thus, while increased model size may lead to better performance, LLMs are still not sensitive to (un)grammaticality the same way as humans are. It seems possible but unlikely that scaling alone can fix this issue. We interpret these results by comparing language learning in vivo and in silico, identifying three critical differences concerning (i) the type of evidence, (ii) the poverty of the stimulus, and (iii) the occurrence of semantic hallucinations due to impenetrable linguistic reference.
Authors: Piyush Jha, Prithwish Jana, Pranavkrishna Suresh, Arnav Arora, Vijay Ganesh
Abstract: Large Language Models (LLMs) have transformed AI but often struggle with tasks that require domain-specific reasoning and logical alignment. Traditional fine-tuning methods do not leverage the vast amount of symbolic domain-knowledge available to us via symbolic reasoning tools (e.g., provers), and are further limited by sparse rewards and unreliable reward models. We introduce Reinforcement Learning via Symbolic Feedback (RLSF), a novel fine-tuning paradigm where symbolic reasoning tools (e.g., solvers, provers, and algebra systems) provide fine-grained feedback to LLMs. RLSF uses poly-sized certificates (e.g., proofs) generated by symbolic tools to identify and correct errors in model outputs, offering token-level guidance without requiring differentiable reasoning systems. This paradigm bridges the gap between symbolic reasoning and LLM fine-tuning, enabling precise alignment with domain-specific constraints while addressing key limitations of traditional reward signals. Via extensive evaluations, we show that our RLSF-based fine-tuning of LLMs outperforms traditional approaches on five different applications (that have some associated logical or domain constraints), namely, program synthesis from natural language pseudo-code to programming language, three chemistry tasks, and solving the Game of 24. A key takeaway is that fine-tuning via RLSF enables relatively smaller LLMs to significantly outperform closed-source models that are orders of magnitude larger.
Authors: Qing Yang, Qiyao Peng, Hongtao Liu, Kai Liu, Bing Qin, Ting Liu
Abstract: Large Language Models (LLMs) have demonstrated exceptional performance across various tasks, with pre-training stage serving as the cornerstone of their capabilities. However, the conventional fixed-length data composition strategy for pre-training presents several practical challenges. When using shorter sequences, documents are often truncated, potentially leading to information loss and affecting the model's ability to capture long-range dependencies. Conversely, longer sequences require concatenation of multiple documents, which can introduce noise and affect the natural document boundaries and semantic coherence as well as require substantial computational overhead. To address these challenges, we first establish three quantitative metrics for evaluating data composition quality: padding ratio, truncation ratio, and concatenation ratio. Building upon these metrics, we propose a novel multi-bucket data composition method that transcends the fixed-length paradigm. Our approach adaptively organizes training data to achieve optimal composition quality as measured by the proposed metrics, offering a more flexible and efficient approach for pre-training. We conduct extensive experiments and the results demonstrate that our proposed method significantly enhances both the efficiency and effectiveness of LLM pre-training.
Authors: Hongcheng Ding, Xuanze Zhao, Ruiting Deng, Shamsul Nahar Abdullah, Deshinta Arrova Dewi, Zixiao Jiang
Abstract: Sentiment analysis plays a crucial role in various domains, such as business intelligence and financial forecasting. Large language models (LLMs) have become a popular paradigm for sentiment analysis, leveraging multi-task learning to address specific tasks concurrently. However, LLMs with fine-tuning for sentiment analysis often underperforms due to the inherent challenges in managing diverse task complexities. Moreover, constant-weight approaches in multi-task learning struggle to adapt to variations in data characteristics, further complicating model effectiveness. To address these issues, we propose a novel multi-task learning framework with a dynamic adaptive optimization (DAO) module. This module is designed as a plug-and-play component that can be seamlessly integrated into existing models, providing an effective and flexible solution for multi-task learning. The key component of the DAO module is dynamic adaptive loss, which dynamically adjusts the weights assigned to different tasks based on their relative importance and data characteristics during training. Sentiment analyses on a standard and customized financial text dataset demonstrate that the proposed framework achieves superior performance. Specifically, this work improves the Mean Squared Error (MSE) and Accuracy (ACC) by 15.58% and 1.24% respectively, compared with previous work.
Authors: Haichuan Hu, Congqing He, Xiaochen Xie, Quanjun Zhang
Abstract: Retrieval-Augmented Generation (RAG) has become a primary technique for mitigating hallucinations in large language models (LLMs). However, incomplete knowledge extraction and insufficient understanding can still mislead LLMs to produce irrelevant or even contradictory responses, which means hallucinations persist in RAG. In this paper, we propose LRP4RAG, a method based on the Layer-wise Relevance Propagation (LRP) algorithm for detecting hallucinations in RAG. Specifically, we first utilize LRP to compute the relevance between the input and output of the RAG generator. We then apply further extraction and resampling to the relevance matrix. The processed relevance data are input into multiple classifiers to determine whether the output contains hallucinations. To the best of our knowledge, this is the first time that LRP has been used for detecting RAG hallucinations, and extensive experiments demonstrate that LRP4RAG outperforms existing baselines.
Authors: Junyoung Lee, Ninad Dixit, Kaustav Chakrabarti, S. Supraja
Abstract: Effective question classification is crucial for AI-driven educational tools, enabling adaptive learning systems to categorize questions by skill area, difficulty level, and competence. It not only supports educational diagnostics and analytics but also enhances complex downstream tasks like information retrieval and question answering by associating questions with relevant categories. Traditional methods, often based on word embeddings and conventional classifiers, struggle to capture the nuanced relationships in question statements, leading to suboptimal performance. We propose a novel approach leveraging graph convolutional networks, named Phrase Question-Graph Convolutional Network (PQ-GCN). Through PQ-GCN, we evaluate the incorporation of phrase-based features to enhance classification performance on question datasets of various domains and characteristics. The proposed method, augmented with phrase-based features, outperform baseline graph-based methods in low-resource settings, and performs competitively against language model-based methods with a fraction of their parameter size. Our findings offer a possible solution for more context-aware, parameter-efficient question classification, bridging the gap between graph neural network research and its educational applications.
Authors: Tianyu Gao, Alexander Wettig, Howard Yen, Danqi Chen
Abstract: We study continued training and supervised fine-tuning (SFT) of a language model (LM) to make effective use of long-context information. We first establish a reliable evaluation protocol to guide model development -- instead of perplexity or simple needle-in-a-haystack (NIAH) tests, we use a broad set of long-context downstream tasks, and we evaluate models after SFT as this better reveals long-context abilities. Supported by our robust evaluations, we run thorough experiments to decide the data mix for continued pre-training, the instruction tuning dataset, and many other design choices such as position extrapolation. We find that (1) code repositories and books are excellent sources of long data, but it is crucial to combine them with high-quality short-context data; (2) training with a sequence length beyond the evaluation length boosts long-context performance; (3) for SFT, using only short instruction datasets yields strong performance on long-context tasks. Our final model, ProLong-8B, which is initialized from Llama-3 and trained on 40B tokens, demonstrates state-of-the-art long-context performance among similarly sized models at a length of 128K. ProLong outperforms Llama-3.1-8B-Instruct on the majority of long-context tasks despite using only 5% as many tokens during long-context training. Additionally, ProLong can effectively process up to 512K tokens, one of the longest context windows of publicly available LMs.
Authors: Robert E. Blackwell, Jon Barry, Anthony G. Cohn
Abstract: Large language models (LLMs) are stochastic, and not all models give deterministic answers, even when setting temperature to zero with a fixed random seed. However, few benchmark studies attempt to quantify uncertainty, partly due to the time and cost of repeated experiments. We use benchmarks designed for testing LLMs' capacity to reason about cardinal directions to explore the impact of experimental repeats on mean score and prediction interval. We suggest a simple method for cost-effectively quantifying the uncertainty of a benchmark score and make recommendations concerning reproducible LLM evaluation.
Authors: Hongcheng Ding, Fuzhen Hu, Ruiting Deng, Xuanze Zhao, Shamsul Nahar Abdullah, Deshinta Arrova Dewi
Abstract: Sentiment analysis has become increasingly important for assessing public opinion and informing decision-making. Large language models (LLMs) have revolutionized this field by capturing nuanced language patterns. However, adapting LLMs to domain-specific sentiment analysis tasks remains challenging due to computational constraints and the need for optimal fine-tuning. To address these challenges, we propose a novel Dynamic Adaptive Rank Space Exploration (DARSE) framework for efficient and effective sentiment analysis using LLMs. DARSE consists of a coarse-grained greedy algorithm to identify the optimal rank range, a fine-grained exploration algorithm to refine rank selection, and a dynamic rank allocation method to determine the optimal rank combination for each LLM layer. Extensive experiments demonstrate that DARSE significantly improves sentiment analysis accuracy, achieving a 15.1% improvement in MSE and a 4.3% improvement in accuracy compared to previous work. Our framework strikes a balance between computational efficiency and model performance, making it a promising approach for sentiment analysis with LLMs.
Authors: Advait Deshmukh, Ashwin Umadi, Dananjay Srinivas, Maria Leonor Pacheco
Abstract: Due to their capacity to acquire world knowledge from large corpora, pre-trained language models (PLMs) are extensively used in ultra-fine entity typing tasks where the space of labels is extremely large. In this work, we explore the limitations of the knowledge acquired by PLMs by proposing a novel heuristic to approximate the pre-training distribution of entities when the pre-training data is unknown. Then, we systematically demonstrate that entity-typing approaches that rely solely on the parametric knowledge of PLMs struggle significantly with entities at the long tail of the pre-training distribution, and that knowledge-infused approaches can account for some of these shortcomings. Our findings suggest that we need to go beyond PLMs to produce solutions that perform well for infrequent entities.
Authors: Anthony Cui, Pranav Nandyalam, Andrew Rufail, Ethan Cheung, Aiden Lei, Kevin Zhu, Sean O'Brien
Abstract: Momentum-Aided Prompt Optimization (MAPO) enhances the efficiency and efficacy of prompt optimization for Large Language Models (LLMs). Building on ProTeGi, MAPO uses positive natural language "gradients" and a momentum-based extension to refine prompts effectively. By tracking gradient history, MAPO avoids local minima and oscillations. It also utilizes beam search and an Upper Confidence Bound (UCB) algorithm for balanced candidate expansion and selection. Benchmark testing shows that MAPO achieves faster convergence time with fewer API calls and higher F1 scores than ProTeGi, proving it as a robust and scalable solution for automated prompt engineering in LLMs.
Authors: Shaden Shaar, Wayne Chen, Maitreyi Chatterjee, Barry Wang, Wenting Zhao, Claire Cardie
Abstract: Most existing work on event extraction has focused on sentence-level texts and presumes the identification of a trigger-span -- a word or phrase in the input that evokes the occurrence of an event of interest. Event arguments are then extracted with respect to the trigger. Indeed, triggers are treated as integral to, and trigger detection as an essential component of, event extraction. In this paper, we provide the first investigation of the role of triggers for the more difficult and much less studied task of document-level event extraction. We analyze their usefulness in multiple end-to-end and pipelined transformer-based event extraction models for three document-level event extraction datasets, measuring performance using triggers of varying quality (human-annotated, LLM-generated, keyword-based, and random). We find that whether or not systems benefit from explicitly extracting triggers depends both on dataset characteristics (i.e. the typical number of events per document) and task-specific information available during extraction (i.e. natural language event schemas). Perhaps surprisingly, we also observe that the mere existence of triggers in the input, even random ones, is important for prompt-based in-context learning approaches to the task.
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: Jiahui Li, Roman Klinger
Abstract: Prompt engineering has made significant contributions to the era of large language models, yet its effectiveness depends on the skills of a prompt author. This paper introduces $\textit{iPrOp}$, a novel interactive prompt optimization approach, to bridge manual prompt engineering and automatic prompt optimization while offering users the flexibility to assess evolving prompts. We aim to provide users with task-specific guidance to enhance human engagement in the optimization process, which is structured through prompt variations, informative instances, predictions generated by large language models along with their corresponding explanations, and relevant performance metrics. This approach empowers users to choose and further refine the prompts based on their individual preferences and needs. It can not only assist non-technical domain experts in generating optimal prompts tailored to their specific tasks or domains, but also enable to study the intrinsic parameters that influence the performance of prompt optimization. The evaluation shows that our approach has the capability to generate improved prompts, leading to enhanced task performance.
Authors: Xinxin Liu, Aaron Thomas, Cheng Zhang, Jianyi Cheng, Yiren Zhao, Xitong Gao
Abstract: Parameter-Efficient Fine-Tuning (PEFT) has gained prominence through low-rank adaptation methods like LoRA. In this paper, we focus on sparsity-based PEFT (SPEFT), which introduces trainable sparse adaptations to the weight matrices in the model, offering greater flexibility in selecting fine-tuned parameters compared to low-rank methods. We conduct the first systematic evaluation of salience metrics for SPEFT, inspired by zero-cost NAS proxies, and identify simple gradient-based metrics is reliable, and results are on par with the best alternatives, offering both computational efficiency and robust performance. Additionally, we compare static and dynamic masking strategies, finding that static masking, which predetermines non-zero entries before training, delivers efficiency without sacrificing performance, while dynamic masking offers no substantial benefits. Across NLP tasks, a simple gradient-based, static SPEFT consistently outperforms other fine-tuning methods for LLMs, providing a simple yet effective baseline for SPEFT. Our work challenges the notion that complexity is necessary for effective PEFT, while our open-source framework establishes a reproducible benchmark for future research, which is available at [https://github.com/0-ml/speft].
Authors: Rohit Saxena, Hao Tang, Frank Keller
Abstract: Training transformer-based encoder-decoder models for long document summarization poses a significant challenge due to the quadratic memory consumption during training. Several approaches have been proposed to extend the input length at test time, but training with these approaches is still difficult, requiring truncation of input documents and causing a mismatch between training and test conditions. In this work, we propose CachED (Gradient $\textbf{Cach}$ing for $\textbf{E}$ncoder-$\textbf{D}$ecoder models), an approach that enables end-to-end training of existing transformer-based encoder-decoder models, using the entire document without truncation. Specifically, we apply non-overlapping sliding windows to input documents, followed by fusion in decoder. During backpropagation, the gradients are cached at the decoder and are passed through the encoder in chunks by re-computing the hidden vectors, similar to gradient checkpointing. In the experiments on long document summarization, we extend BART to CachED BART, processing more than 500K tokens during training and achieving superior performance without using any additional parameters.
Authors: Tianyu Gao, Alexander Wettig, Luxi He, Yihe Dong, Sadhika Malladi, Danqi Chen
Abstract: The vast diversity of styles, domains, and quality levels present in language model pre-training corpora is essential in developing general model capabilities, but efficiently learning and deploying the correct behaviors exemplified in each of these heterogeneous data sources is challenging. To address this, we propose a new method, termed Metadata Conditioning then Cooldown (MeCo), to incorporate additional learning cues during pre-training. MeCo first provides metadata (e.g., URLs like www$.$wikipedia$.$org) alongside the text during training and later uses a cooldown phase with only the standard text, thereby enabling the model to function normally even without metadata. MeCo significantly accelerates pre-training across different model scales (600M to 8B parameters) and training sources (C4, RefinedWeb, and DCLM). For instance, a 1.6B language model trained with MeCo matches the downstream task performance of standard pre-training while using 33% less data. Additionally, MeCo enables us to steer language models by conditioning the inference prompt on either real or fabricated metadata that encodes the desired properties of the output: for example, prepending wikipedia$.$org to reduce harmful generations or factquizmaster$.$com (fabricated) to improve common knowledge task performance. We also demonstrate that MeCo is compatible with different types of metadata, such as model-generated topics. MeCo is remarkably simple, adds no computational overhead, and demonstrates promise in producing more capable and steerable language models.
Authors: Steven H. Wang, Maksim Zubkov, Kexin Fan, Sarah Harrell, Yuyang Sun, Wei Chen, Andreas Plesner, Roger Wattenhofer
Abstract: Information retrieval, specifically contract clause retrieval, is foundational to contract drafting because lawyers rarely draft contracts from scratch; instead, they locate and revise the most relevant precedent. We introduce the Atticus Clause Retrieval Dataset (ACORD), the first retrieval benchmark for contract drafting fully annotated by experts. ACORD focuses on complex contract clauses such as Limitation of Liability, Indemnification, Change of Control, and Most Favored Nation. It includes 114 queries and over 126,000 query-clause pairs, each ranked on a scale from 1 to 5 stars. The task is to find the most relevant precedent clauses to a query. The bi-encoder retriever paired with pointwise LLMs re-rankers shows promising results. However, substantial improvements are still needed to effectively manage the complex legal work typically undertaken by lawyers. As the first retrieval benchmark for contract drafting annotated by experts, ACORD can serve as a valuable IR benchmark for the NLP community.
Authors: Sadegh Mahdavi, Muchen Li, Kaiwen Liu, Christos Thrampoulidis, Leonid Sigal, Renjie Liao
Abstract: Advances in Large Language Models (LLMs) have sparked interest in their ability to solve Olympiad-level math problems. However, the training and evaluation of these models are constrained by the limited size and quality of available datasets, as creating large-scale data for such advanced problems requires extensive effort from human experts. In addition, current benchmarks are prone to contamination, leading to unreliable evaluations. In this paper, we present an automated pipeline that leverages the rich resources of the Art of Problem Solving (AoPS) forum, which predominantly features Olympiad-level problems and community-driven solutions. Using open-source LLMs, we develop a method to extract question-answer pairs from the forum, resulting in AoPS-Instruct, a dataset of more than 600,000 high-quality QA pairs. Our experiments demonstrate that fine-tuning LLMs on AoPS-Instruct improves their reasoning abilities across various benchmarks. Moreover, we build an automatic pipeline that introduces LiveAoPSBench, an evolving evaluation set with timestamps, derived from the latest forum data, providing a contamination-resistant benchmark for assessing LLM performance. Notably, we observe a significant decline in LLM performance over time, suggesting their success on older examples may stem from pre-training exposure rather than true reasoning ability. Our work presents a scalable approach to creating and maintaining large-scale, high-quality datasets for advanced math reasoning, offering valuable insights into the capabilities and limitations of LLMs in this domain. Our benchmark and code is available at https://github.com/DSL-Lab/aops
Authors: S. M. Yousuf Iqbal Tomal, Abdullah Al Shafin, Debojit Bhattacharjee, MD. Khairul Amin, Rafiad Sadat Shahir
Abstract: Recent advances in quantum computing have opened new pathways for enhancing deep learning architectures, particularly in domains characterized by high-dimensional and context-rich data such as natural language processing (NLP). In this work, we present a hybrid classical-quantum Transformer model that integrates a quantum-enhanced attention mechanism into the standard classical architecture. By embedding token representations into a quantum Hilbert space via parameterized variational circuits and exploiting entanglement-aware kernel similarities, the model captures complex semantic relationships beyond the reach of conventional dot-product attention. We demonstrate the effectiveness of this approach across diverse NLP benchmarks, showing improvements in both efficiency and representational capacity. The results section reveal that the quantum attention layer yields globally coherent attention maps and more separable latent features, while requiring comparatively fewer parameters than classical counterparts. These findings highlight the potential of quantum-classical hybrid models to serve as a powerful and resource-efficient alternative to existing attention mechanisms in NLP.
Authors: Xiang Liu, Zhenheng Tang, Peijie Dong, Zeyu Li, Yue Liu, Bo Li, Xuming Hu, Xiaowen Chu
Abstract: Large Language Models (LLMs) require significant GPU memory when processing long texts, with the key value (KV) cache consuming up to 70\% of total memory during inference. Although existing compression methods reduce memory by evaluating the importance of individual tokens, they overlook critical semantic relationships between tokens, resulting in fragmented context and degraded performance. We introduce ChunkKV, which fundamentally reimagines KV cache compression by treating semantic chunks - rather than isolated tokens - as basic compression units. This approach preserves complete linguistic structures and contextual integrity, ensuring that essential meaning is retained even under aggressive compression. Our innovation includes a novel layer-wise index reuse technique that exploits the higher cross-layer similarity of preserved indices in ChunkKV, reducing computational overhead and improving throughput by 26.5\%. Comprehensive evaluations on challenging benchmarks: LongBench, Needle-In-A-HayStack, GSM8K, and JailbreakV demonstrate that ChunkKV outperforms state-of-the-art methods by up to 8.7\% in precision while maintaining the same compression ratio. These results confirm that semantic-aware compression significantly enhances both efficiency and performance for long-context LLM inference, providing a simple yet effective solution to the memory bottleneck problem.
Authors: Yichi Zhang, Siyuan Zhang, Yao Huang, Zeyu Xia, Zhengwei Fang, Xiao Yang, Ranjie Duan, Dong Yan, Yinpeng Dong, Jun Zhu
Abstract: Ensuring the safety and harmlessness of Large Language Models (LLMs) has become equally critical as their performance in applications. However, existing safety alignment methods typically suffer from safety-performance trade-offs and the susceptibility to jailbreak attacks, primarily due to their reliance on direct refusals for malicious queries. In this paper, we propose STAIR, a novel framework that integrates SafeTy Alignment with Itrospective Reasoning. We enable LLMs to identify safety risks through step-by-step analysis by self-improving chain-of-thought (CoT) reasoning with safety awareness. STAIR first equips the model with a structured reasoning capability and then advances safety alignment via iterative preference optimization on step-level reasoning data generated using our newly proposed Safety-Informed Monte Carlo Tree Search (SI-MCTS). We further train a process reward model on this data to guide test-time searches for improved responses. Extensive experiments show that STAIR effectively mitigates harmful outputs while better preserving helpfulness, compared to instinctive alignment strategies. With test-time scaling, STAIR achieves a safety performance comparable to Claude-3.5 against popular jailbreak attacks. Relevant resources in this work are available at https://github.com/thu-ml/STAIR.
Authors: Xuejiao Zhao, Siyan Liu, Su-Yin Yang, Chunyan Miao
Abstract: Retrieval-augmented generation (RAG) is a well-suited technique for retrieving privacy-sensitive Electronic Health Records (EHR). It can serve as a key module of the healthcare copilot, helping reduce misdiagnosis for healthcare practitioners and patients. However, the diagnostic accuracy and specificity of existing heuristic-based RAG models used in the medical domain are inadequate, particularly for diseases with similar manifestations. This paper proposes MedRAG, a RAG model enhanced by knowledge graph (KG)-elicited reasoning for the medical domain that retrieves diagnosis and treatment recommendations based on manifestations. MedRAG systematically constructs a comprehensive four-tier hierarchical diagnostic KG encompassing critical diagnostic differences of various diseases. These differences are dynamically integrated with similar EHRs retrieved from an EHR database, and reasoned within a large language model. This process enables more accurate and specific decision support, while also proactively providing follow-up questions to enhance personalized medical decision-making. MedRAG is evaluated on both a public dataset DDXPlus and a private chronic pain diagnostic dataset (CPDD) collected from Tan Tock Seng Hospital, and its performance is compared against various existing RAG methods. Experimental results show that, leveraging the information integration and relational abilities of the KG, our MedRAG provides more specific diagnostic insights and outperforms state-of-the-art models in reducing misdiagnosis rates. Our code will be available at https://github.com/SNOWTEAM2023/MedRAG
Authors: Hongbin Na, Yining Hua, Zimu Wang, Tao Shen, Beibei Yu, Lilin Wang, Wei Wang, John Torous, Ling Chen
Abstract: Mental health is increasingly critical in contemporary healthcare, with psychotherapy demanding dynamic, context-sensitive interactions that traditional NLP methods struggle to capture. Large Language Models (LLMs) offer significant potential for addressing this gap due to their ability to handle extensive context and multi-turn reasoning. This review introduces a conceptual taxonomy dividing psychotherapy into interconnected stages--assessment, diagnosis, and treatment--to systematically examine LLM advancements and challenges. Our comprehensive analysis reveals imbalances in current research, such as a focus on common disorders, linguistic biases, fragmented methods, and limited theoretical integration. We identify critical challenges including capturing dynamic symptom fluctuations, overcoming linguistic and cultural biases, and ensuring diagnostic reliability. Highlighting future directions, we advocate for continuous multi-stage modeling, real-time adaptive systems grounded in psychological theory, and diversified research covering broader mental disorders and therapeutic approaches, aiming toward more holistic and clinically integrated psychotherapy LLMs systems.
Authors: Jonathan Jordan, Sherzod Hakimov, David Schlangen
Abstract: Large Language Models (LLMs) serve not only as chatbots but as key components in agent systems, where their common-sense knowledge significantly impacts performance as language-based planners for situated or embodied action. We assess LLMs' incremental learning (based on feedback from the environment), and controlled in-context learning abilities using a text-based environment. We introduce challenging yet interesting set of experiments to test i) how agents can incrementally solve tasks related to every day objects in typical rooms in a house where each of them are discovered by interacting within the environment, ii) controlled in-context learning abilities and efficiency of agents by providing short info about locations of objects and rooms to check how faster the task can be solved, and finally iii) using synthetic pseudo-English words to gauge how well LLMs are at inferring meaning of unknown words from environmental feedback. Results show that larger commercial models have a substantial gap in performance compared to open-weight but almost all models struggle with the synthetic words experiments.
Authors: Zhitao He, Zijun Liu, Peng Li, Yi R Fung, Ming Yan, Ji Zhang, Fei Huang, Yang Liu
Abstract: LLM-based agents have made significant advancements in interactive environments, such as mobile operations and web browsing, and other domains beyond computer using. Current multi-agent systems universally excel in performance, compared to single agents, but struggle with generalization across environments due to predefined roles and inadequate strategies for generalizing language agents. The challenge of achieving both strong performance and good generalization has hindered the progress of multi-agent systems for interactive environments. To address these issues, we propose CollabUIAgents, a multi-agent reinforcement learning framework with a novel multi-agent credit re-assignment (CR) strategy, assigning process rewards with LLMs rather than environment-specific rewards and learning with synthesized preference data, in order to foster generalizable, collaborative behaviors among the role-free agents' policies. Empirical results show that our framework improves both performance and cross-environment generalizability of multi-agent systems. Moreover, our 7B-parameter system achieves results on par with or exceed strong closed-source models, and the LLM that guides the CR. We also provide insights in using granular CR rewards effectively for environment generalization, and accommodating trained LLMs in multi-agent systems.
Authors: Yaohua Tang, Zhicheng Hu, Kun Cheng, Fan Mo, Qiheng Lv, Hua Wang, Zhi Chen
Abstract: The increasing context window size in large language models (LLMs) has improved their ability to handle complex, long-text tasks. However, as the conversation rounds continue, it is required to store a large amount of KV cache in GPU memory, which significantly affects the efficiency and even availability of the model serving systems. This paper analyzes dialogue data from real users on the granularity of round and discovers that the LLM inference manifests a watershed layer, after which the distribution of round-level attention shows notable similarity. Based on this, we propose Round Attention - a novel round-level attention mechanism that selectively processes the KV cache of top-k relevant rounds, where k is dynamically determined through the attention matrix in the watershed layer. Theoretical analysis demonstrates that our method reduces memory usage by 54\% to 82\%, while experimental results confirm that loading sparse critical-round KV cache maintains answer accuracy without performance degradation.
Authors: Zhuo Chen, Xinyu Wang, Yong Jiang, Zhen Zhang, Xinyu Geng, Pengjun Xie, Fei Huang, Kewei Tu
Abstract: Despite the advancements made in Visual Large Language Models (VLLMs), like text Large Language Models (LLMs), they have limitations in addressing questions that require real-time information or are knowledge-intensive. Indiscriminately adopting Retrieval Augmented Generation (RAG) techniques is an effective yet expensive way to enable models to answer queries beyond their knowledge scopes. To mitigate the dependence on retrieval and simultaneously maintain, or even improve, the performance benefits provided by retrieval, we propose a method to detect the knowledge boundary of VLLMs, allowing for more efficient use of techniques like RAG. Specifically, we propose a method with two variants that fine-tunes a VLLM on an automatically constructed dataset for boundary identification. Experimental results on various types of Visual Question Answering datasets show that our method successfully depicts a VLLM's knowledge boundary based on which we are able to reduce indiscriminate retrieval while maintaining or improving the performance. In addition, we show that the knowledge boundary identified by our method for one VLLM can be used as a surrogate boundary for other VLLMs. Code will be released at https://github.com/Chord-Chen-30/VLLM-KnowledgeBoundary
URLs: https://github.com/Chord-Chen-30/VLLM-KnowledgeBoundary
Authors: Karl Audun Borgersen, Morten Goodwin
Abstract: For consumer usage of locally deployed LLMs, the GGUF format and k\_quantization are invaluable tools for maintaining the performance of the original model while reducing it to sizes deployable with consumer-grade hardware. The number of bits dedicated to each weight from the original model is reduced based on how important they are thought to be during model inference. This importance is arrived at through the application of an 'importance matrix'-a relatively small text document meant to be representative of the LLM's standard use-cases. In the vast majority of quants available online, this document is primarily written in English. It was therefore an open question whether performance on English language tasks was preserved through the sacrifice of multilingual performance and whether it can be preserved with alternate importance matrices. This article investigates these hypotheses by quantizing Llama3.3 70B on importance matrices written in three languages (English, Norwegian, and Malayalam) and evaluating them on the MixEval dataset in both English and Norwegian. All experiments related to yielded non-significant results indicating that current quantization practices do not disproportionately harm multilingual performance.
Authors: Xinyi He, Yihao Liu, Mengyu Zhou, Yeye He, Haoyu Dong, Shi Han, Zejian Yuan, Dongmei Zhang
Abstract: Tabular data are crucial in many fields and their understanding by large language models (LLMs) under high parameter efficiency paradigm is important. However, directly applying parameter-efficient fine-tuning (PEFT) techniques to tabular tasks presents significant challenges, particularly in terms of better table serialization and the representation of two-dimensional structured information within a one-dimensional sequence. To address this, we propose TableLoRA, a module designed to improve LLMs' understanding of table structure during PEFT. It incorporates special tokens for serializing tables with special token encoder and uses 2D LoRA to encode low-rank information on cell positions. Experiments on four tabular-related datasets demonstrate that TableLoRA consistently outperforms vanilla LoRA and surpasses various table encoding methods tested in control experiments. These findings reveal that TableLoRA, as a table-specific LoRA, enhances the ability of LLMs to process tabular data effectively, especially in low-parameter settings, demonstrating its potential as a robust solution for handling table-related tasks.
Authors: Tsunehiko Tanaka, Edgar Simo-Serra
Abstract: Game Description Generation (GDG) is the task of generating a game description written in a Game Description Language (GDL) from natural language text. Previous studies have explored generation methods leveraging the contextual understanding capabilities of Large Language Models (LLMs); however, accurately reproducing the game features of the game descriptions remains a challenge. In this paper, we propose reinforcement learning-based fine-tuning of LLMs for GDG (RLGDG). Our training method simultaneously improves grammatical correctness and fidelity to game concepts by introducing both grammar rewards and concept rewards. Furthermore, we adopt a two-stage training strategy where Reinforcement Learning (RL) is applied following Supervised Fine-Tuning (SFT). Experimental results demonstrate that our proposed method significantly outperforms baseline methods using SFT alone. Our code is available at https://github.com/tsunehiko/rlgdg
Authors: Yang Tian, Zheng Lu, Mingqi Gao, Zheng Liu, Bo Zhao
Abstract: Fully comprehending scientific papers by machines reflects a high level of Artificial General Intelligence, requiring the ability to reason across fragmented and heterogeneous sources of information, presenting a complex and practically significant challenge. While Vision-Language Models (VLMs) have made remarkable strides in various tasks, particularly those involving reasoning with evidence source from single image or text page, their ability to use cross-source information for reasoning remains an open problem. This work presents MMCR, a high-difficulty benchmark designed to evaluate VLMs' capacity for reasoning with cross-source information from scientific papers. The benchmark comprises 276 high-quality questions, meticulously annotated by humans across 7 subjects and 10 task types. Experiments with 18 VLMs demonstrate that cross-source reasoning presents a substantial challenge for existing models. Notably, even the top-performing model, GPT-4o, achieved only 48.55% overall accuracy, with only 20% accuracy in multi-table comprehension tasks, while the second-best model, Qwen2.5-VL-72B, reached 39.86% overall accuracy. Furthermore, we investigated the impact of the Chain-of-Thought (CoT) technique on cross-source reasoning and observed a detrimental effect on small models, whereas larger models demonstrated substantially enhanced performance. These results highlight the pressing need to develop VLMs capable of effectively utilizing cross-source information for reasoning.
Authors: Ren\'e Peinl, Vincent Tischler
Abstract: Similar to LLMs, the development of vision language models is mainly driven by English datasets and models trained in English and Chinese language, whereas support for other languages, even those considered high-resource languages such as German, remains significantly weaker. In this work we present an analysis of open-weight VLMs on factual knowledge in the German and English language. We disentangle the image-related aspects from the textual ones by analyzing accu-racy with jury-as-a-judge in both prompt languages and images from German and international contexts. We found that for celebrities and sights, VLMs struggle because they are lacking visual cognition of German image contents. For animals and plants, the tested models can often correctly identify the image contents ac-cording to the scientific name or English common name but fail in German lan-guage. Cars and supermarket products were identified equally well in English and German images across both prompt languages.
Authors: Haoming Yang, Ke Ma, Xiaojun Jia, Yingfei Sun, Qianqian Xu, Qingming Huang
Abstract: Despite the remarkable performance of Large Language Models (LLMs), they remain vulnerable to jailbreak attacks, which can compromise their safety mechanisms. Existing studies often rely on brute-force optimization or manual design, failing to uncover potential risks in real-world scenarios. To address this, we propose a novel jailbreak attack framework, ICRT, inspired by heuristics and biases in human cognition. Leveraging the simplicity effect, we employ cognitive decomposition to reduce the complexity of malicious prompts. Simultaneously, relevance bias is utilized to reorganize prompts, enhancing semantic alignment and inducing harmful outputs effectively. Furthermore, we introduce a ranking-based harmfulness evaluation metric that surpasses the traditional binary success-or-failure paradigm by employing ranking aggregation methods such as Elo, HodgeRank, and Rank Centrality to comprehensively quantify the harmfulness of generated content. Experimental results show that our approach consistently bypasses mainstream LLMs' safety mechanisms and generates high-risk content, providing insights into jailbreak attack risks and contributing to stronger defense strategies.
Authors: Jingcheng Niu, Xingdi Yuan, Tong Wang, Hamidreza Saghir, Amir H. Abdi
Abstract: We observe a novel phenomenon, contextual entrainment, across a wide range of language models (LMs) and prompt settings, providing a new mechanistic perspective on how LMs become distracted by ``irrelevant'' contextual information in the input prompt. Specifically, LMs assign significantly higher logits (or probabilities) to any tokens that have previously appeared in the context prompt, even for random tokens. This suggests that contextual entrainment is a mechanistic phenomenon, occurring independently of the relevance or semantic relation of the tokens to the question or the rest of the sentence. We find statistically significant evidence that the magnitude of contextual entrainment is influenced by semantic factors. Counterfactual prompts have a greater effect compared to factual ones, suggesting that while contextual entrainment is a mechanistic phenomenon, it is modulated by semantic factors. We hypothesise that there is a circuit of attention heads -- the entrainment heads -- that corresponds to the contextual entrainment phenomenon. Using a novel entrainment head discovery method based on differentiable masking, we identify these heads across various settings. When we ``turn off'' these heads, i.e., set their outputs to zero, the effect of contextual entrainment is significantly attenuated, causing the model to generate output that capitulates to what it would produce if no distracting context were provided. Our discovery of contextual entrainment, along with our investigation into LM distraction via the entrainment heads, marks a key step towards the mechanistic analysis and mitigation of the distraction problem.
Authors: Chengyu Wang, Junbing Yan, Wenrui Cai, Yuanhao Yue, Jun Huang
Abstract: In this paper, we present EasyDistill, a comprehensive toolkit designed for effective black-box and white-box knowledge distillation (KD) of large language models (LLMs). Our framework offers versatile functionalities, including data synthesis, supervised fine-tuning, ranking optimization, and reinforcement learning techniques specifically tailored for KD scenarios. The toolkit accommodates KD functionalities for both System 1 (fast, intuitive) and System 2 (slow, analytical) models. With its modular design and user-friendly interface, EasyDistill empowers researchers and industry practitioners to seamlessly experiment with and implement state-of-the-art KD strategies for LLMs. In addition, EasyDistill provides a series of robust distilled models and KD-based industrial solutions developed by us, along with the corresponding open-sourced datasets, catering to a variety of use cases. Furthermore, we describe the seamless integration of EasyDistill into Alibaba Cloud's Platform for AI (PAI). Overall, the EasyDistill toolkit makes advanced KD techniques for LLMs more accessible and impactful within the NLP community.
Authors: Zhitao He, Sandeep Polisetty, Zhiyuan Fan, Yuchen Huang, Shujin Wu, Yi R. Fung
Abstract: In recent years, multimodal large language models (MLLMs) have made significant progress but continue to face inherent challenges in multimodal reasoning, which requires multi-level (e.g., perception, reasoning) and multi-granular (e.g., multi-step reasoning chain) advanced inferencing. Prior work on estimating model confidence tends to focus on the overall response for training and calibration, but fails to assess confidence in each reasoning step, leading to undesirable hallucination snowballing. In this work, we present MMBoundary, a novel framework that advances the knowledge boundary awareness of MLLMs through reasoning step confidence calibration. To achieve this, we propose to incorporate complementary textual and cross-modal self-rewarding signals to estimate confidence at each step of the MLLM reasoning process. In addition to supervised fine-tuning MLLM on this set of self-rewarded confidence estimation signal for initial confidence expression warm-up, we introduce a reinforcement learning stage with multiple reward functions for further aligning model knowledge and calibrating confidence at each reasoning step, enhancing reasoning chain self-correction. Empirical results show that MMBoundary significantly outperforms existing methods across diverse domain datasets and metrics, achieving an average of 7.5% reduction in multimodal confidence calibration errors and up to 8.3% improvement in task performance.
Authors: Nikita Martynov, Anastasia Mordasheva, Dmitriy Gorbetskiy, Danil Astafurov, Ulyana Isaeva, Elina Basyrova, Sergey Skachkov, Victoria Berestova, Nikolay Ivanov, Valeriia Zanina, Alena Fenogenova
Abstract: We introduce POLLUX, a comprehensive open-source benchmark designed to evaluate the generative capabilities of large language models (LLMs) in Russian. Our main contribution is a novel evaluation methodology that enhances the interpretability of LLM assessment. For each task type, we define a set of detailed criteria and develop a scoring protocol where models evaluate responses and provide justifications for their ratings. This enables transparent, criteria-driven evaluation beyond traditional resource-consuming, side-by-side human comparisons. POLLUX includes a detailed, fine-grained taxonomy of 35 task types covering diverse generative domains such as code generation, creative writing, and practical assistant use cases, totaling 2,100 manually crafted and professionally authored prompts. Each task is categorized by difficulty (easy/medium/hard), with experts constructing the dataset entirely from scratch. We also release a family of LLM-as-a-Judge (7B and 32B) evaluators trained for nuanced assessment of generative outputs. This approach provides scalable, interpretable evaluation and annotation tools for model development, effectively replacing costly and less precise human judgments.
Authors: Dana Lupsa, Sanda-Maria Avram, Radu Lupsa
Abstract: This study addresses the problem of authorship attribution for Romanian texts using the ROST corpus, a standard benchmark in the field. We systematically evaluate six machine learning techniques: Support Vector Machine (SVM), Logistic Regression (LR), k-Nearest Neighbors (k-NN), Decision Trees (DT), Random Forests (RF), and Artificial Neural Networks (ANN), employing character n-gram features for classification. Among these, the ANN model achieved the highest performance, including perfect classification in four out of fifteen runs when using 5-gram features. These results demonstrate that lightweight, interpretable character n-gram approaches can deliver state-of-the-art accuracy for Romanian authorship attribution, rivaling more complex methods. Our findings highlight the potential of simple stylometric features in resource, constrained or under-studied language settings.
Authors: Hao Li, Viktor Schlegel, Yizheng Sun, Riza Batista-Navarro, Goran Nenadic
Abstract: Argument Mining (AM), a critical subfield of Natural Language Processing (NLP), focuses on extracting argumentative structures from text. The advent of Large Language Models (LLMs) has profoundly transformed AM, enabling advanced in-context learning, prompt-based generation, and robust cross-domain adaptability. This survey systematically synthesizes recent advancements in LLM-driven AM. We provide a concise review of foundational theories and annotation frameworks, alongside a meticulously curated catalog of datasets. A key contribution is our comprehensive taxonomy of AM subtasks, elucidating how contemporary LLM techniques -- such as prompting, chain-of-thought reasoning, and retrieval augmentation -- have reconfigured their execution. We further detail current LLM architectures and methodologies, critically assess evaluation practices, and delineate pivotal challenges including long-context reasoning, interpretability, and annotation bottlenecks. Conclusively, we highlight emerging trends and propose a forward-looking research agenda for LLM-based computational argumentation, aiming to strategically guide researchers in this rapidly evolving domain.
Authors: Yingji Zhang, Danilo S. Carvalho, Andr\'e Freitas
Abstract: Integrating compositional and symbolic properties into current distributional semantic spaces can enhance the interpretability, controllability, compositionality, and generalisation capabilities of Transformer-based auto-regressive language models (LMs). In this survey, we offer a novel perspective on latent space geometry through the lens of compositional semantics, a direction we refer to as \textit{semantic representation learning}. This direction enables a bridge between symbolic and distributional semantics, helping to mitigate the gap between them. We review and compare three mainstream autoencoder architectures-Variational AutoEncoder (VAE), Vector Quantised VAE (VQVAE), and Sparse AutoEncoder (SAE)-and examine the distinctive latent geometries they induce in relation to semantic structure and interpretability.
Authors: Kaixiang Zhang, Justine Zhang, Cristian Danescu-Niculescu-Mizil
Abstract: An intrinsic aspect of every conversation is the way talk-time is shared between multiple speakers. Conversations can be balanced, with each speaker claiming a similar amount of talk-time, or imbalanced when one talks disproportionately. Such overall distributions are the consequence of continuous negotiations between the speakers throughout the conversation: who should be talking at every point in time, and for how long? In this work we introduce a computational framework for quantifying both the conversation-level distribution of talk-time between speakers, as well as the lower-level dynamics that lead to it. We derive a typology of talk-time sharing dynamics structured by several intuitive axes of variation. By applying this framework to a large dataset of video-chats between strangers, we confirm that, perhaps unsurprisingly, different conversation-level distributions of talk-time are perceived differently by speakers, with balanced conversations being preferred over imbalanced ones, especially by those who end up talking less. Then we reveal that -- even when they lead to the same level of overall balance -- different types of talk-time sharing dynamics are perceived differently by the participants, highlighting the relevance of our newly introduced typology. Finally, we discuss how our framework offers new tools to designers of computer-mediated communication platforms, for both human-human and human-AI communication.
Authors: Hongcheng Ding, Xuanze Zhao, Ruiting Deng, Shamsul Nahar Abdullah, Deshinta Arrova Dewi
Abstract: Accurate forecasting of the EUR/USD exchange rate is crucial for investors, businesses, and policymakers. This paper proposes a novel framework, IUS, that integrates unstructured textual data from news and analysis with structured data on exchange rates and financial indicators to enhance exchange rate prediction. The IUS framework employs large language models for sentiment polarity scoring and exchange rate movement classification of texts. These textual features are combined with quantitative features and input into a Causality-Driven Feature Generator. An Optuna-optimized Bi-LSTM model is then used to forecast the EUR/USD exchange rate. Experiments demonstrate that the proposed method outperforms benchmark models, reducing MAE by 10.69% and RMSE by 9.56% compared to the best performing baseline. Results also show the benefits of data fusion, with the combination of unstructured and structured data yielding higher accuracy than structured data alone. Furthermore, feature selection using the top 12 important quantitative features combined with the textual features proves most effective. The proposed IUS framework and Optuna-Bi-LSTM model provide a powerful new approach for exchange rate forecasting through multi-source data integration.
Authors: Zhen Qin, Zhaomin Wu, Bingsheng He, Shuiguang Deng
Abstract: Instruction tuning is a crucial step in improving the responsiveness of pretrained large language models (LLMs) to human instructions. Federated learning (FL) helps to exploit the use of vast private instruction data from clients, becoming popular for LLM tuning by improving data diversity. Existing federated tuning simply consumes all local data, causing excessive computational overhead and overfitting to local data, while centralized data-efficient solutions are not suitable for FL due to privacy concerns. This work presents FedHDS, a federated data-efficient instruction tuning approach, which tunes LLMs with a representative subset of edge-side data. It reduces the data redundancy at both intra- and inter-client levels without sharing raw data. Experiments with various LLMs, datasets and partitions show that FedHDS improves Rouge-L on unseen tasks by an average of 10.72% over the SOTA full-data federated instruction tuning methods, while using less than 1.5% of the data samples, improving training efficiency by up to tens of times.
Authors: Xiang Li, Feng Ruan, Huiyuan Wang, Qi Long, Weijie J. Su
Abstract: Watermarking has offered an effective approach to distinguishing text generated by large language models (LLMs) from human-written text. However, the pervasive presence of human edits on LLM-generated text dilutes watermark signals, thereby significantly degrading detection performance of existing methods. In this paper, by modeling human edits through mixture model detection, we introduce a new method in the form of a truncated goodness-of-fit test for detecting watermarked text under human edits, which we refer to as Tr-GoF. We prove that the Tr-GoF test achieves optimality in robust detection of the Gumbel-max watermark in a certain asymptotic regime of substantial text modifications and vanishing watermark signals. Importantly, Tr-GoF achieves this optimality \textit{adaptively} as it does not require precise knowledge of human edit levels or probabilistic specifications of the LLMs, in contrast to the optimal but impractical (Neyman--Pearson) likelihood ratio test. Moreover, we establish that the Tr-GoF test attains the highest detection efficiency rate in a certain regime of moderate text modifications. In stark contrast, we show that sum-based detection rules, as employed by existing methods, fail to achieve optimal robustness in both regimes because the additive nature of their statistics is less resilient to edit-induced noise. Finally, we demonstrate the competitive and sometimes superior empirical performance of the Tr-GoF test on both synthetic data and open-source LLMs in the OPT and LLaMA families.
Authors: Qiushi Sun, Kanzhi Cheng, Zichen Ding, Chuanyang Jin, Yian Wang, Fangzhi Xu, Zhenyu Wu, Chengyou Jia, Liheng Chen, Zhoumianze Liu, Ben Kao, Guohao Li, Junxian He, Yu Qiao, Zhiyong Wu
Abstract: Graphical User Interface (GUI) agents powered by Vision-Language Models (VLMs) have demonstrated human-like computer control capability. Despite their utility in advancing digital automation, a critical bottleneck persists: collecting high-quality trajectory data for training. Common practices for collecting such data rely on human supervision or synthetic data generation through executing pre-defined tasks, which are either resource-intensive or unable to guarantee data quality. Moreover, these methods suffer from limited data diversity and significant gaps between synthetic data and real-world environments. To address these challenges, we propose OS-Genesis, a novel GUI data synthesis pipeline that reverses the conventional trajectory collection process. Instead of relying on pre-defined tasks, OS-Genesis enables agents first to perceive environments and perform step-wise interactions, then retrospectively derive high-quality tasks to enable trajectory-level exploration. A trajectory reward model is then employed to ensure the quality of the generated trajectories. We demonstrate that training GUI agents with OS-Genesis significantly improves their performance on highly challenging online benchmarks. In-depth analysis further validates OS-Genesis's efficiency and its superior data quality and diversity compared to existing synthesis methods. Our codes, data, and checkpoints are available at https://qiushisun.github.io/OS-Genesis-Home/.
Authors: Karthik Mohan, Pengyu Chen
Abstract: In this paper, we describe our systems in which the objective is to determine whether a given news article could be considered as hyperpartisan. Hyperpartisan news is news that takes an extremely polarized political standpoint with an intention of creating political divide among the public. We attempted several approaches, including n-grams, sentiment analysis, as well as sentence and document representation using pre-tained ELMo. Our best system using pre-trained ELMo with Bidirectional LSTM achieved an accuracy of 83% through 10-fold cross-validation without much hyperparameter tuning.
Authors: Shiji Zhao, Ranjie Duan, Fengxiang Wang, Chi Chen, Caixin Kang, Shouwei Ruan, Jialing Tao, YueFeng Chen, Hui Xue, Xingxing Wei
Abstract: Multimodal Large Language Models (MLLMs) have achieved impressive performance and have been put into practical use in commercial applications, but they still have potential safety mechanism vulnerabilities. Jailbreak attacks are red teaming methods that aim to bypass safety mechanisms and discover MLLMs' potential risks. Existing MLLMs' jailbreak methods often bypass the model's safety mechanism through complex optimization methods or carefully designed image and text prompts. Despite achieving some progress, they have a low attack success rate on commercial closed-source MLLMs. Unlike previous research, we empirically find that there exists a Shuffle Inconsistency between MLLMs' comprehension ability and safety ability for the shuffled harmful instruction. That is, from the perspective of comprehension ability, MLLMs can understand the shuffled harmful text-image instructions well. However, they can be easily bypassed by the shuffled harmful instructions from the perspective of safety ability, leading to harmful responses. Then we innovatively propose a text-image jailbreak attack named SI-Attack. Specifically, to fully utilize the Shuffle Inconsistency and overcome the shuffle randomness, we apply a query-based black-box optimization method to select the most harmful shuffled inputs based on the feedback of the toxic judge model. A series of experiments show that SI-Attack can improve the attack's performance on three benchmarks. In particular, SI-Attack can obviously improve the attack success rate for commercial MLLMs such as GPT-4o or Claude-3.5-Sonnet.
Authors: Ahmed Heakl, Abdullah Sohail, Mukul Ranjan, Rania Hossam, Ghazi Shazan Ahmad, Mohamed El-Geish, Omar Maher, Zhiqiang Shen, Fahad Khan, Salman Khan
Abstract: With the growing adoption of Retrieval-Augmented Generation (RAG) in document processing, robust text recognition has become increasingly critical for knowledge extraction. While OCR (Optical Character Recognition) for English and other languages benefits from large datasets and well-established benchmarks, Arabic OCR faces unique challenges due to its cursive script, right-to-left text flow, and complex typographic and calligraphic features. We present KITAB-Bench, a comprehensive Arabic OCR benchmark that fills the gaps in current evaluation systems. Our benchmark comprises 8,809 samples across 9 major domains and 36 sub-domains, encompassing diverse document types including handwritten text, structured tables, and specialized coverage of 21 chart types for business intelligence. Our findings show that modern vision-language models (such as GPT-4o, Gemini, and Qwen) outperform traditional OCR approaches (like EasyOCR, PaddleOCR, and Surya) by an average of 60% in Character Error Rate (CER). Furthermore, we highlight significant limitations of current Arabic OCR models, particularly in PDF-to-Markdown conversion, where the best model Gemini-2.0-Flash achieves only 65% accuracy. This underscores the challenges in accurately recognizing Arabic text, including issues with complex fonts, numeral recognition errors, word elongation, and table structure detection. This work establishes a rigorous evaluation framework that can drive improvements in Arabic document analysis methods and bridge the performance gap with English OCR technologies.
Authors: Arnav Kumar Jain, Gonzalo Gonzalez-Pumariega, Wayne Chen, Alexander M Rush, Wenting Zhao, Sanjiban Choudhury
Abstract: We address the problem of code generation from multi-turn execution feedback. Existing methods either generate code without feedback or use complex, hierarchical reinforcement learning to optimize multi-turn rewards. We propose a simple yet scalable approach, $\mu$Code, that solves multi-turn code generation using only single-step rewards. Our key insight is that code generation is a one-step recoverable MDP, where the correct code can be recovered from any intermediate code state in a single turn. $\mu$Code iteratively trains both a generator to provide code solutions conditioned on multi-turn execution feedback and a verifier to score the newly generated code. Experimental evaluations show that our approach achieves significant improvements over the state-of-the-art baselines. We provide analysis of the design choices of the reward models and policy, and show the efficacy of $\mu$Code at utilizing the execution feedback. Our code is available at https://github.com/portal-cornell/muCode.
Authors: Seyed Pouyan Mousavi Davoudi, Amin Gholami Davodi, Alireza Amiri-Margavi, 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: Xi Zhu, Haochen Xue, Ziwei Zhao, Wujiang Xu, Jingyuan Huang, Minghao Guo, Qifan Wang, Kaixiong Zhou, Yongfeng Zhang
Abstract: Text-Attributed Graphs (TAGs), where each node is associated with text descriptions, are ubiquitous in real-world scenarios. They typically exhibit distinctive structure and domain-specific knowledge, motivating the development of a Graph Foundation Model (GFM) that generalizes across diverse graphs and tasks. Despite large efforts to integrate Large Language Models (LLMs) and Graph Neural Networks (GNNs) for TAGs, existing approaches suffer from decoupled architectures with two-stage alignment, limiting their synergistic potential. Even worse, existing methods assign out-of-vocabulary (OOV) tokens to graph nodes, leading to graph-specific semantics, token explosion, and incompatibility with task-oriented prompt templates, which hinders cross-graph and cross-task transferability. To address these challenges, we propose PromptGFM, a versatile GFM for TAGs grounded in graph vocabulary learning. PromptGFM comprises two key components: (1) Graph Understanding Module, which explicitly prompts LLMs to replicate the finest GNN workflow within the text space, facilitating seamless GNN-LLM integration and elegant graph-text alignment; (2) Graph Inference Module, which establishes a language-based graph vocabulary ensuring expressiveness, transferability, and scalability, enabling readable instructions for LLM fine-tuning. Extensive experiments demonstrate our superiority and transferability across diverse graphs and tasks. The code is available at this: https://github.com/agiresearch/PromptGFM.
Authors: Can Zheng, Jiguang He, Guofa Cai, Zitong Yu, Chung G. Kang
Abstract: In this paper, we propose BeamLLM, a vision-aided millimeter-wave (mmWave) beam prediction framework leveraging large language models (LLMs) to address the challenges of high training overhead and latency in mmWave communication systems. By combining computer vision (CV) with LLMs' cross-modal reasoning capabilities, the framework extracts user equipment (UE) positional features from RGB images and aligns visual-temporal features with LLMs' semantic space through reprogramming techniques. Evaluated on a realistic vehicle-to-infrastructure (V2I) scenario, the proposed method achieves 61.01% top-1 accuracy and 97.39% top-3 accuracy in standard prediction tasks, significantly outperforming traditional deep learning models. In few-shot prediction scenarios, the performance degradation is limited to 12.56% (top-1) and 5.55% (top-3) from time sample 1 to 10, demonstrating superior prediction capability.
Authors: Qiushi Sun, Zhoumianze Liu, Chang Ma, Zichen Ding, Fangzhi Xu, Zhangyue Yin, Haiteng Zhao, Zhenyu Wu, Kanzhi Cheng, Zhaoyang Liu, Jianing Wang, Qintong Li, Xiangru Tang, Tianbao Xie, Xiachong Feng, Xiang Li, Ben Kao, Wenhai Wang, Biqing Qi, Lingpeng Kong, Zhiyong Wu
Abstract: Large Language Models (LLMs) have extended their impact beyond Natural Language Processing, substantially fostering the development of interdisciplinary research. Recently, various LLM-based agents have been developed to assist scientific discovery progress across multiple aspects and domains. Among these, computer-using agents, capable of interacting with operating systems as humans do, are paving the way to automated scientific problem-solving and addressing routines in researchers' workflows. Recognizing the transformative potential of these agents, we introduce ScienceBoard, which encompasses two complementary contributions: (i) a realistic, multi-domain environment featuring dynamic and visually rich scientific workflows with integrated professional software, where agents can autonomously interact via different interfaces to accelerate complex research tasks and experiments; and (ii) a challenging benchmark of 169 high-quality, rigorously validated real-world tasks curated by humans, spanning scientific-discovery workflows in domains such as biochemistry, astronomy, and geoinformatics. Extensive evaluations of agents with state-of-the-art backbones (e.g., GPT-4o, Claude 3.7, UI-TARS) show that, despite some promising results, they still fall short of reliably assisting scientists in complex workflows, achieving only a 15% overall success rate. In-depth analysis further provides valuable insights for addressing current agent limitations and more effective design principles, paving the way to build more capable agents for scientific discovery. Our code, environment, and benchmark are at https://qiushisun.github.io/ScienceBoard-Home/.
Authors: Ren\'e Peinl, Vincent Tischler
Abstract: This paper introduces a novel benchmark dataset designed to evaluate the capabilities of Vision Language Models (VLMs) on tasks that combine visual reasoning with subject-specific background knowledge in the German language. In contrast to widely used English-language benchmarks that often rely on artificially difficult or decontextualized problems, this dataset draws from real middle school curricula across nine domains including mathematics, history, biology, and religion. The benchmark includes over 2,000 open-ended questions grounded in 486 images, ensuring that models must integrate visual interpretation with factual reasoning rather than rely on superficial textual cues. We evaluate thirteen state-of-the-art open-weight VLMs across multiple dimensions, including domain-specific accuracy and performance on adversarial crafted questions. Our findings reveal that even the strongest models achieve less than 45% overall accuracy, with particularly poor performance in music, mathematics, and adversarial settings. Furthermore, the results indicate significant discrepancies between success on popular benchmarks and real-world multimodal understanding. We conclude that middle school-level tasks offer a meaningful and underutilized avenue for stress-testing VLMs, especially in non-English contexts. The dataset and evaluation protocol serve as a rigorous testbed to better understand and improve the visual and linguistic reasoning capabilities of future AI systems.