Authors: Atharva Bhargude, Ishan Gonehal, Chandler Haney, Dave Yoon, Kevin Zhu, Aaron Sandoval, Sean O'Brien, Kaustubh Vinnakota
Abstract: Phishing attacks represent a significant cybersecurity threat, necessitating adaptive detection techniques. This study explores few-shot Adaptive Linguistic Prompting (ALP) in detecting phishing webpages through the multimodal capabilities of state-of-the-art large language models (LLMs) such as GPT-4o and Gemini 1.5 Pro. ALP is a structured semantic reasoning method that guides LLMs to analyze textual deception by breaking down linguistic patterns, detecting urgency cues, and identifying manipulative diction commonly found in phishing content. By integrating textual, visual, and URL-based analysis, we propose a unified model capable of identifying sophisticated phishing attempts. Our experiments demonstrate that ALP significantly enhances phishing detection accuracy by guiding LLMs through structured reasoning and contextual analysis. The findings highlight the potential of ALP-integrated multimodal LLMs to advance phishing detection frameworks, achieving an F1-score of 0.93, surpassing traditional approaches. These results establish a foundation for more robust, interpretable, and adaptive linguistic-based phishing detection systems using LLMs.
Authors: Keito Inoshita, Rushia Harada
Abstract: In the field of emotion recognition, the development of high-performance models remains a challenge due to the scarcity of high-quality, diverse emotional datasets. Emotional expressions are inherently subjective, shaped by individual personality traits, socio-cultural backgrounds, and contextual factors, making large-scale, generalizable data collection both ethically and practically difficult. To address this issue, we introduce PersonaGen, a novel framework for generating emotionally rich text using a Large Language Model (LLM) through multi-stage persona-based conditioning. PersonaGen constructs layered virtual personas by combining demographic attributes, socio-cultural backgrounds, and detailed situational contexts, which are then used to guide emotion expression generation. We conduct comprehensive evaluations of the generated synthetic data, assessing semantic diversity through clustering and distributional metrics, human-likeness via LLM-based quality scoring, realism through comparison with real-world emotion corpora, and practical utility in downstream emotion classification tasks. Experimental results show that PersonaGen significantly outperforms baseline methods in generating diverse, coherent, and discriminative emotion expressions, demonstrating its potential as a robust alternative for augmenting or replacing real-world emotional datasets.
Authors: Rafiq Kamel, Filippo Guerranti, Simon Geisler, Stephan G\"unnemann
Abstract: Large Language Models (LLMs) are increasingly applied to tasks involving structured inputs such as graphs. Abstract Meaning Representations (AMRs), which encode rich semantics as directed graphs, offer a rigorous testbed for evaluating LLMs on text generation from such structures. Yet, current methods often arbitrarily linearize AMRs, discarding key structural cues, or rely on architectures incompatible with standard LLMs. We introduce SAFT, a structure-aware fine-tuning approach that injects graph topology into pretrained LLMs without architectural changes. We compute direction-sensitive positional encodings from the magnetic Laplacian of transformed AMRs and project them into the embedding space of the LLM. While possibly applicable to any graph-structured inputs, we focus on AMR-to-text generation as a representative and challenging benchmark. SAFT sets a new state-of-the-art on AMR 3.0 with a 3.5 BLEU improvement over baselines. Gains scale with graph complexity, highlighting the value of structure-aware representations in enhancing LLM performance. SAFT offers a general and effective pathway for bridging structured data and language models.
Authors: Chandrashekar Muniyappa, Sirisha Velampalli
Abstract: In today\'s digital world, fake news is spreading with immense speed. Its a significant concern to address. In this work, we addressed that challenge using novel graph based approach. We took dataset from Kaggle that contains real and fake news articles. To test our approach we incorporated recent covid-19 related news articles that contains both genuine and fake news that are relevant to this problem. This further enhances the dataset as well instead of relying completely on the original dataset. We propose a contextual graph-based approach to detect fake news articles. We need to convert news articles into appropriate schema, so we leverage Natural Language Processing (NLP) techniques to transform news articles into contextual graph structures. We then apply the Minimum Description Length (MDL)-based Graph-Based Anomaly Detection (GBAD) algorithm for graph mining. Graph-based methods are particularly effective for handling rich contextual data, as they enable the discovery of complex patterns that traditional query-based or statistical techniques might overlook. Our proposed approach identifies normative patterns within the dataset and subsequently uncovers anomalous patterns that deviate from these established norms.
Authors: Kundeshwar Pundalik, Piyush Sawarkar, Nihar Sahoo, Abhishek Shinde, Prateek Chanda, Vedant Goswami, Ajay Nagpal, Atul Singh, Viraj Thakur, Vijay Dewane, Aamod Thakur, Bhargav Patel, Smita Gautam, Bhagwan Panditi, Shyam Pawar, Madhav Kotcha, Suraj Racha, Saral Sureka, Pankaj Singh, Rishi Bal, Rohit Saluja, Ganesh Ramakrishnan
Abstract: Large Language Models (LLMs) have emerged as powerful general-purpose reasoning systems, yet their development remains dominated by English-centric data, architectures, and optimization paradigms. This exclusionary design results in structural under-representation of linguistically diverse regions such as India, where over 20 official languages and 100+ dialects coexist alongside phenomena like code-switching and diglossia. We introduce PARAM-1, a 2.9B parameter decoder-only, text-only language model trained from scratch with an explicit architectural and linguistic focus on Indian diversity. PARAM-1 is trained on a bilingual dataset consisting of only Hindi and English, constructed with a strong focus on fact-rich, high-quality content. It is guided by three core principles: equitable representation of Indic languages through a 25% corpus allocation; tokenization fairness via a SentencePiece tokenizer adapted to Indian morphological structures; and culturally aligned evaluation benchmarks across IndicQA, code-mixed reasoning, and socio-linguistic robustness tasks. By embedding diversity at the pretraining level-rather than deferring it to post-hoc alignment-PARAM-1 offers a design-first blueprint for equitable foundation modeling. Our results demonstrate that it serves as both a competent general-purpose model and a robust baseline for India-centric applications.
Authors: Emil H\"aglund, Johanna Bj\"orklund
Abstract: We improve the extraction of insights from customer reviews by restructuring the topic modelling pipeline to operate on opinion units - distinct statements that include relevant text excerpts and associated sentiment scores. Prior work has demonstrated that such units can be reliably extracted using large language models. The result is a heightened performance of the subsequent topic modeling, leading to coherent and interpretable topics while also capturing the sentiment associated with each topic. By correlating the topics and sentiments with business metrics, such as star ratings, we can gain insights on how specific customer concerns impact business outcomes. We present our system's implementation, use cases, and advantages over other topic modeling and classification solutions. We also evaluate its effectiveness in creating coherent topics and assess methods for integrating topic and sentiment modalities for accurate star-rating prediction.
Authors: Xuanqi Gao, Weipeng Jiang, Juan Zhai, Shiqing Ma, Siyi Xie, Xinyang Yin, Chao Shen
Abstract: The advent of neural machine translation (NMT) has revolutionized cross-lingual communication, yet preserving stylistic nuances remains a significant challenge. While existing approaches often require parallel corpora for style preservation, we introduce Babel, a novel framework that enhances stylistic fidelity in NMT using only monolingual corpora. Babel employs two key components: (1) a style detector based on contextual embeddings that identifies stylistic disparities between source and target texts, and (2) a diffusion-based style applicator that rectifies stylistic inconsistencies while maintaining semantic integrity. Our framework integrates with existing NMT systems as a post-processing module, enabling style-aware translation without requiring architectural modifications or parallel stylistic data. Extensive experiments on five diverse domains (law, literature, scientific writing, medicine, and educational content) demonstrate Babel's effectiveness: it identifies stylistic inconsistencies with 88.21% precision and improves stylistic preservation by 150% while maintaining a high semantic similarity score of 0.92. Human evaluation confirms that translations refined by Babel better preserve source text style while maintaining fluency and adequacy.
Authors: Cheng-Ting Chou, George Liu, Jessica Sun, Cole Blondin, Kevin Zhu, Vasu Sharma, Sean O'Brien
Abstract: Deterministically controlling the target generation language of large multilingual language models (LLMs) remains a fundamental challenge, particularly in zero-shot settings where neither explicit language prompts nor fine-tuning are available. In this work, we investigate whether sparse autoencoder (SAE) features, previously shown to correlate with interpretable model behaviors, can be leveraged to steer the generated language of LLMs during inference. Leveraging pretrained SAEs on the residual streams of Gemma-2B and Gemma-9B, we identify features whose activations differ most significantly between English and four target languages: Chinese, Japanese, Spanish, and French. By modifying just a single SAE feature at one transformer layer, we achieve controlled language shifts with up to 90\% success, as measured by FastText language classification, while preserving semantic fidelity according to LaBSE (Language-Agnostic BERT Sentence Embedding) similarity. Our analysis reveals that language steering is most effective in mid-to-late transformer layers and is amplified by specific attention heads disproportionately associated with language-sensitive SAE features. These results demonstrate the promise of sparse feature steering as a lightweight and interpretable mechanism for controllable multilingual generation.
Authors: Nur A Zarin Nishat, Andrea Coletta, Luigi Bellomarini, Kossi Amouzouvi, Jens Lehmann, Sahar Vahdati
Abstract: Large language models like GPT-4, Gemini, and Claude have transformed natural language processing (NLP) tasks such as question answering, dialogue generation, summarization, and so forth; yet their susceptibility to hallucination stands as one of the major challenges. Among numerous approaches to overcome this challenge, integration of Knowledge Graphs (KGs) into language models has emerged as a promising solution as it provides structured, reliable, domain-specific, and up-to-date external information to the language models. In this paper, we introduce ALIGNed-LLM, a simple yet effective approach to improve language models' factuality via a lean strategy to infuse KGs into the latent space of language models inspired by LLaVA where visual and textual information is infused. We use embeddings from a pre-trained Knowledge Graph Embedding (KGE) model, such as TransE, and a trainable projection layer to align entity and text embeddings. This alignment enables the language model to distinguish between similar entities improving factual grounding and reducing hallucination. We tested our approach on three popular questions-answering benchmark datasets alongside language models of varying sizes, showing significant improvement. Furthermore, we applied our approach to a real-world financial use case from a large central bank in Europe, which demands high accuracy and precision, demonstrating a substantial improvement of the LLM answers.
Authors: Liang Lin, Zhihao Xu, Xuehai Tang, Shi Liu, Biyu Zhou, Fuqing Zhu, Jizhong Han, Songlin Hu
Abstract: The safety of large language models (LLMs) has garnered significant research attention. In this paper, we argue that previous empirical studies demonstrate LLMs exhibit a propensity to trust information from authoritative sources, such as academic papers, implying new possible vulnerabilities. To verify this possibility, a preliminary analysis is designed to illustrate our two findings. Based on this insight, a novel jailbreaking method, Paper Summary Attack (\llmname{PSA}), is proposed. It systematically synthesizes content from either attack-focused or defense-focused LLM safety paper to construct an adversarial prompt template, while strategically infilling harmful query as adversarial payloads within predefined subsections. Extensive experiments show significant vulnerabilities not only in base LLMs, but also in state-of-the-art reasoning model like Deepseek-R1. PSA achieves a 97\% attack success rate (ASR) on well-aligned models like Claude3.5-Sonnet and an even higher 98\% ASR on Deepseek-R1. More intriguingly, our work has further revealed diametrically opposed vulnerability bias across different base models, and even between different versions of the same model, when exposed to either attack-focused or defense-focused papers. This phenomenon potentially indicates future research clues for both adversarial methodologies and safety alignment.Code is available at https://github.com/233liang/Paper-Summary-Attack
Authors: Siqi Shen, Mehar Singh, Lajanugen Logeswaran, Moontae Lee, Honglak Lee, Rada Mihalcea
Abstract: There has been extensive research on assessing the value orientation of Large Language Models (LLMs) as it can shape user experiences across demographic groups. However, several challenges remain. First, while the Multiple Choice Question (MCQ) setting has been shown to be vulnerable to perturbations, there is no systematic comparison of probing methods for value probing. Second, it is unclear to what extent the probed values capture in-context information and reflect models' preferences for real-world actions. In this paper, we evaluate the robustness and expressiveness of value representations across three widely used probing strategies. We use variations in prompts and options, showing that all methods exhibit large variances under input perturbations. We also introduce two tasks studying whether the values are responsive to demographic context, and how well they align with the models' behaviors in value-related scenarios. We show that the demographic context has little effect on the free-text generation, and the models' values only weakly correlate with their preference for value-based actions. Our work highlights the need for a more careful examination of LLM value probing and awareness of its limitations.
Authors: Matilde Marcolli, Robert C. Berwick
Abstract: We provide a mathematical argument showing that, given a representation of lexical items as functions (wavelets, for instance) in some function space, it is possible to construct a faithful representation of arbitrary syntactic objects in the same function space. This space can be endowed with a commutative non-associative semiring structure built using the second Renyi entropy. The resulting representation of syntactic objects is compatible with the magma structure. The resulting set of functions is an algebra over an operad, where the operations in the operad model circuits that transform the input wave forms into a combined output that encodes the syntactic structure. The action of Merge on workspaces is faithfully implemented as action on these circuits, through a coproduct and a Hopf algebra Markov chain. The results obtained here provide a constructive argument showing the theoretical possibility of a neurocomputational realization of the core computational structure of syntax. We also present a particular case of this general construction where this type of realization of Merge is implemented as a cross frequency phase synchronization on sinusoidal waves. This also shows that Merge can be expressed in terms of the successor function of a semiring, thus clarifying the well known observation of its similarities with the successor function of arithmetic.
Authors: Mohamed Achref Ben Ammar, Mohamed Taha Bennani
Abstract: The analysis of conversational dynamics has gained increasing importance with the rise of large language model-based systems, which interact with users across diverse contexts. In this work, we propose a novel computational framework for constructing conversational graphs that capture the flow and structure of loosely organized dialogues, referred to as quasi-patterned conversations. We introduce the Filter & Reconnect method, a novel graph simplification technique that minimizes noise while preserving semantic coherence and structural integrity of conversational graphs. Through comparative analysis, we demonstrate that the use of large language models combined with our graph simplification technique has resulted in semantic metric S increasing by a factor of 2.06 compared to previous approaches while simultaneously enforcing a tree-like structure with 0 {\delta}-hyperbolicity, ensuring optimal clarity in conversation modeling. This work provides a computational method for analyzing large-scale dialogue datasets, with practical applications related to monitoring automated systems such as chatbots, dialogue management tools, and user behavior analytics.
Authors: Feng Chen, Weizhe Xu, Changye Li, Serguei Pakhomov, Alex Cohen, Simran Bhola, Sandy Yin, Sunny X Tang, Michael Mackinley, Lena Palaniyappan, Dror Ben-Zeev, Trevor Cohen
Abstract: Formal thought disorder (FTD), a hallmark of schizophrenia spectrum disorders, manifests as incoherent speech and poses challenges for clinical assessment. Traditional clinical rating scales, though validated, are resource-intensive and lack scalability. Automated speech analysis with automatic speech recognition (ASR) allows for objective quantification of linguistic and temporal features of speech, offering scalable alternatives. The use of utterance timestamps in ASR captures pause dynamics, which are thought to reflect the cognitive processes underlying speech production. However, the utility of integrating these ASR-derived features for assessing FTD severity requires further evaluation. This study integrates pause features with semantic coherence metrics across three datasets: naturalistic self-recorded diaries (AVH, n = 140), structured picture descriptions (TOPSY, n = 72), and dream narratives (PsyCL, n = 43). We evaluated pause related features alongside established coherence measures, using support vector regression (SVR) to predict clinical FTD scores. Key findings demonstrate that pause features alone robustly predict the severity of FTD. Integrating pause features with semantic coherence metrics enhanced predictive performance compared to semantic-only models, with integration of independent models achieving correlations up to \r{ho} = 0.649 and AUC = 83.71% for severe cases detection (TOPSY, with best \r{ho} = 0.584 and AUC = 79.23% for semantic-only models). The performance gains from semantic and pause features integration held consistently across all contexts, though the nature of pause patterns was dataset-dependent. These findings suggest that frameworks combining temporal and semantic analyses provide a roadmap for refining the assessment of disorganized speech and advance automated speech analysis in psychosis.
Authors: Kirill Borodin, Nikita Vasiliev, Vasiliy Kudryavtsev, Maxim Maslov, Mikhail Gorodnichev, Oleg Rogov, Grach Mkrtchian
Abstract: Russian speech synthesis presents distinctive challenges, including vowel reduction, consonant devoicing, variable stress patterns, homograph ambiguity, and unnatural intonation. This paper introduces Balalaika, a novel dataset comprising more than 2,000 hours of studio-quality Russian speech with comprehensive textual annotations, including punctuation and stress markings. Experimental results show that models trained on Balalaika significantly outperform those trained on existing datasets in both speech synthesis and enhancement tasks. We detail the dataset construction pipeline, annotation methodology, and results of comparative evaluations.
Authors: Sergio E. Zanotto, Segun Aroyehun
Abstract: The rapid advancements in large language models (LLMs) have significantly improved their ability to generate natural language, making texts generated by LLMs increasingly indistinguishable from human-written texts. While recent research has primarily focused on using LLMs to classify text as either human-written and machine-generated texts, our study focus on characterizing these texts using a set of linguistic features across different linguistic levels such as morphology, syntax, and semantics. We select a dataset of human-written and machine-generated texts spanning 8 domains and produced by 11 different LLMs. We calculate different linguistic features such as dependency length and emotionality and we use them for characterizing human-written and machine-generated texts along with different sampling strategies, repetition controls and model release date. Our statistical analysis reveals that human-written texts tend to exhibit simpler syntactic structures and more diverse semantic content. Furthermore, we calculate the variability of our set of features across models and domains. Both human and machine texts show stylistic diversity across domains, with humans displaying greater variation in our features. Finally, we apply style embeddings to further test variability among human-written and machine-generated texts. Notably, newer models output text that is similarly variable, pointing to an homogenization of machine-generated texts.
Authors: Shanbo Cheng, Yu Bao, Qian Cao, Luyang Huang, Liyan Kang, Zhicheng Liu, Yu Lu, Wenhao Zhu, Zhichao Huang, Tao Li, Sitong Liu, Ningxin Peng, Shuaijie She, Lu Xu, Nuo Xu, Sen Yang, Runsheng Yu, Yiming Yu, Liehao Zou, Hang Li, Lu Lu, Yuxuan Wang, Yonghui Wu
Abstract: Multilingual translation stands as a challenging task for large language models (LLMs) to handle intricate language patterns and stilted translations that arise in automated translations. In this paper, we introduce Seed-X, a family of open-source LLMs comprising instruct and reasoning models, pushing the limits of translation capability with 7B parameter size. The base model is pre-trained on a diverse, high-quality dataset encompassing both monolingual and bilingual content across 28 languages, harnessing the full potential of multilingual data. The instruct model is then finetuned to translate by Chain-of-Thought (CoT) reasoning and further enhanced through reinforcement learning (RL) to achieve better generalization across diverse language pairs. Seed-X achieves performance comparable to leading closed-source models, including Gemini-2.5 and GPT-4o, across 28 languages, and significantly outperforms larger open-source models in both automatic metrics and human evaluations. We share the best practices through our optimization process, and make the parameter public available for advancing translation research and applications.
Authors: Teerapong Panboonyuen
Abstract: Integrating large language models into specialized domains like healthcare presents unique challenges, including domain adaptation and limited labeled data. We introduce CU-ICU, a method for customizing unsupervised instruction-finetuned language models for ICU datasets by leveraging the Text-to-Text Transfer Transformer (T5) architecture. CU-ICU employs a sparse fine-tuning approach that combines few-shot prompting with selective parameter updates, enabling efficient adaptation with minimal supervision. Our evaluation across critical ICU tasks--early sepsis detection, mortality prediction, and clinical note generation--demonstrates that CU-ICU consistently improves predictive accuracy and interpretability over standard fine-tuning methods. Notably, CU-ICU achieves up to a 15% increase in sepsis detection accuracy and a 20% enhancement in generating clinically relevant explanations while updating fewer than 1% of model parameters in its most efficient configuration. These results establish CU-ICU as a scalable, low-overhead solution for delivering accurate and interpretable clinical decision support in real-world ICU environments.
Authors: Woo-Chan Kim, Ji-Hoon Park, Seong-Whan Lee
Abstract: Large language models (LLMs) have demonstrated state-of-the-art performance across a wide range of natural language processing tasks. However, high-performing models are typically accessible only via APIs, incurring substantial inference costs. Cascade methods address this by initially employing a cheaper model and escalating to a stronger one only when necessary. Nevertheless, existing cascade approaches struggle to select a reliable representative response and assess the overall reliability of free-form outputs, as they rely on exact text matching. To overcome these limitations, we propose Keyword-inspired Cascade (KiC), a novel framework for cost-efficient free-form text generation. KiC identifies the most representative answer among multiple outputs from a weaker model and evaluates the semantic alignment of other responses with it. Based on the degree of alignment, KiC determines whether to accept the weaker model's output or escalate to a stronger model. Experiments on three free-form text generation benchmarks show that KiC achieves 97.53 percent of GPT-4's accuracy while reducing API costs by 28.81 percent on average, and even outperforms GPT-4 in a specific benchmark.
Authors: Haoyang Li, Zhanchao Xu, Yiming Li, Xuejia Chen, Darian Li, Anxin Tian, Qingfa Xiao, Cheng Deng, Jun Wang, Qing Li, Lei Chen, Mingxuan Yuan
Abstract: Multi-turn dialogues are essential in many real-world applications of large language models, such as chatbots and virtual assistants. As conversation histories become longer, existing large language models face increasing computational and memory challenges, which hinder their ability to provide efficient and responsive interactions. Most current acceleration methods either compress the context or optimize key value caching, but they often rely on fixed or position-based heuristics that do not adapt well to the dynamic and unpredictable patterns found in actual multi-turn conversations. In this paper, we present LoopServe, an adaptive dual-phase inference acceleration framework for large language models in multi-turn dialogues. LoopServe introduces two main innovations. First, it performs online sparsification during the prefilling phase by dynamically selecting the most important parts of the attention matrix for each new input. Second, it uses progressive key value compression during decoding by adaptively maintaining a relevant and efficient cache based on the most recently generated output tokens. We also propose a \href{https://huggingface.co/datasets/TreeAILab/Multi-turn_Long-context_Benchmark_for_LLMs}{new benchmark} with eleven multi-turn datasets that reflect realistic query positions and conversational dependencies. Extensive experiments demonstrate that LoopServe consistently achieves superior effectiveness compared to existing baselines and significantly accelerates LLM inference across a wide range of long-context dialogue tasks.
URLs: https://huggingface.co/datasets/TreeAILab/Multi-turn_Long-context_Benchmark_for_LLMs
Authors: Cedric Waterschoot, Nava Tintarev, Francesco Barile
Abstract: Large Language Models (LLMs) are increasingly being implemented as joint decision-makers and explanation generators for Group Recommender Systems (GRS). In this paper, we evaluate these recommendations and explanations by comparing them to social choice-based aggregation strategies. Our results indicate that LLM-generated recommendations often resembled those produced by Additive Utilitarian (ADD) aggregation. However, the explanations typically referred to averaging ratings (resembling but not identical to ADD aggregation). Group structure, uniform or divergent, did not impact the recommendations. Furthermore, LLMs regularly claimed additional criteria such as user or item similarity, diversity, or used undefined popularity metrics or thresholds. Our findings have important implications for LLMs in the GRS pipeline as well as standard aggregation strategies. Additional criteria in explanations were dependent on the number of ratings in the group scenario, indicating potential inefficiency of standard aggregation methods at larger item set sizes. Additionally, inconsistent and ambiguous explanations undermine transparency and explainability, which are key motivations behind the use of LLMs for GRS.
Authors: Guillaume Zambrano
Abstract: This study examines the role of human judges in legal decision-making by using machine learning to predict child physical custody outcomes in French appellate courts. Building on the legal realism-formalism debate, we test whether individual judges' decision-making patterns significantly influence case outcomes, challenging the assumption that judges are neutral variables that apply the law uniformly. To ensure compliance with French privacy laws, we implement a strict pseudonymization process. Our analysis uses 18,937 living arrangements rulings extracted from 10,306 cases. We compare models trained on individual judges' past rulings (specialist models) with a judge-agnostic model trained on aggregated data (generalist models). The prediction pipeline is a hybrid approach combining large language models (LLMs) for structured feature extraction and ML models for outcome prediction (RF, XGB and SVC). Our results show that specialist models consistently achieve higher predictive accuracy than the general model, with top-performing models reaching F1 scores as high as 92.85%, compared to the generalist model's 82.63% trained on 20x to 100x more samples. Specialist models capture stable individual patterns that are not transferable to other judges. In-Domain and Cross-Domain validity tests provide empirical support for legal realism, demonstrating that judicial identity plays a measurable role in legal outcomes. All data and code used will be made available.
Authors: Maluna Menke, Thilo Hagendorff
Abstract: Large Language Models (LLMs) frequently reproduce the gender- and sexual-identity prejudices embedded in their training corpora, leading to outputs that marginalize LGBTQIA+ users. Hence, reducing such biases is of great importance. To achieve this, we evaluate two parameter-efficient fine-tuning (PEFT) techniques - Low-Rank Adaptation (LoRA) and soft-prompt tuning - as lightweight alternatives to full-model fine-tuning for mitigating such biases. Using the WinoQueer benchmark, we quantify bias in three open-source LLMs and observe baseline bias scores reaching up to 98 (out of 100) across a range of queer identities defined by gender and/or sexual orientation, where 50 would indicate neutrality. Fine-tuning with LoRA (< 0.1% additional parameters) on a curated QueerNews corpus reduces those scores by up to 50 points and raises neutrality from virtually 0% to as much as 36%. Soft-prompt tuning (10 virtual tokens) delivers only marginal improvements. These findings show that LoRA can deliver meaningful fairness gains with minimal computation. We advocate broader adoption of community-informed PEFT, the creation of larger queer-authored corpora, and richer evaluation suites beyond WinoQueer, coupled with ongoing audits to keep LLMs inclusive.
Authors: Palash Nandi, Maithili Joshi, Tanmoy Chakraborty
Abstract: Language models are highly sensitive to prompt formulations - small changes in input can drastically alter their output. This raises a critical question: To what extent can prompt sensitivity be exploited to generate inapt content? In this paper, we investigate how discrete components of prompt design influence the generation of inappropriate content in Visual Language Models (VLMs). Specifically, we analyze the impact of three key factors on successful jailbreaks: (a) the inclusion of detailed visual information, (b) the presence of adversarial examples, and (c) the use of positively framed beginning phrases. Our findings reveal that while a VLM can reliably distinguish between benign and harmful inputs in unimodal settings (text-only or image-only), this ability significantly degrades in multimodal contexts. Each of the three factors is independently capable of triggering a jailbreak, and we show that even a small number of in-context examples (as few as three) can push the model toward generating inappropriate outputs. Furthermore, we propose a framework that utilizes a skip-connection between two internal layers of the VLM, which substantially increases jailbreak success rates, even when using benign images. Finally, we demonstrate that memes, often perceived as humorous or harmless, can be as effective as toxic visuals in eliciting harmful content, underscoring the subtle and complex vulnerabilities of VLMs.
Authors: Enhao Cheng, Shoujia Zhang, Jianhua Yin, Xuemeng Song, Tian Gan, Liqiang Nie
Abstract: Short text clustering has become increasingly important with the popularity of social media like Twitter, Google+, and Facebook. Existing methods can be broadly categorized into two paradigms: topic model-based approaches and deep representation learning-based approaches. This task is inherently challenging due to the sparse, large-scale, and high-dimensional characteristics of the short text data. Furthermore, the computational intensity required by representation learning significantly increases the running time. To address these issues, we propose a collapsed Gibbs Sampling algorithm for the Dirichlet Multinomial Mixture model (GSDMM), which effectively handles the sparsity and high dimensionality of short texts while identifying representative words for each cluster. Based on several aspects of GSDMM that warrant further refinement, we propose an improved approach, GSDMM+, designed to further optimize its performance. GSDMM+ reduces initialization noise and adaptively adjusts word weights based on entropy, achieving fine-grained clustering that reveals more topic-related information. Additionally, strategic cluster merging is employed to refine clustering granularity, better aligning the predicted distribution with the true category distribution. We conduct extensive experiments, comparing our methods with both classical and state-of-the-art approaches. The experimental results demonstrate the efficiency and effectiveness of our methods. The source code for our model is publicly available at https://github.com/chehaoa/VEMC.
Authors: Hosein Azarbonyad, Zi Long Zhu, Georgios Cheirmpos, Zubair Afzal, Vikrant Yadav, Georgios Tsatsaronis
Abstract: When deciding to read an article or incorporate it into their research, scholars often seek to quickly identify and understand its main ideas. In this paper, we aim to extract these key concepts and contributions from scientific articles in the form of Question and Answer (QA) pairs. We propose two distinct approaches for generating QAs. The first approach involves selecting salient paragraphs, using a Large Language Model (LLM) to generate questions, ranking these questions by the likelihood of obtaining meaningful answers, and subsequently generating answers. This method relies exclusively on the content of the articles. However, assessing an article's novelty typically requires comparison with the existing literature. Therefore, our second approach leverages a Knowledge Graph (KG) for QA generation. We construct a KG by fine-tuning an Entity Relationship (ER) extraction model on scientific articles and using it to build the graph. We then employ a salient triplet extraction method to select the most pertinent ERs per article, utilizing metrics such as the centrality of entities based on a triplet TF-IDF-like measure. This measure assesses the saliency of a triplet based on its importance within the article compared to its prevalence in the literature. For evaluation, we generate QAs using both approaches and have them assessed by Subject Matter Experts (SMEs) through a set of predefined metrics to evaluate the quality of both questions and answers. Our evaluations demonstrate that the KG-based approach effectively captures the main ideas discussed in the articles. Furthermore, our findings indicate that fine-tuning the ER extraction model on our scientific corpus is crucial for extracting high-quality triplets from such documents.
Authors: Lizhi Ma, Tong Zhao, Shuai Zhang, Nirui Song, Hongliang He, Anqi Li, Ran Feng, Huachuan Qiu, Jingsong Ma, Zhenzhong Lan
Abstract: This study explores the relationship between linguistic expressions and psychological states of depression and anxiety within Chinese psycho-counseling interactions, focusing specifically on the usage of first-person singular pronouns and negative emotional words. Utilizing a corpus derived from 735 online counseling sessions, the analysis employed a general linear mixed-effect model to assess linguistic patterns quantified by the Linguistic Inquiry and Word Count (LIWC) software. Results indicate a significant positive correlation between the frequency of negative emotional words and the severity of both depressive and anxious states among clients. However, contrary to prior findings predominantly derived from English-language contexts, the usage frequency of first-person singular pronouns did not vary significantly with the clients' psychological conditions. These outcomes are discussed within the framework of cultural distinctions between collectivist Chinese contexts and individualistic Western settings, as well as the interactive dynamics unique to psycho-counseling conversations. The findings highlight the nuanced influence of cultural and conversational contexts on language use in mental health communications, providing insights into psycholinguistic markers relevant to therapeutic practices in Chinese-speaking populations.
Authors: Eitan Wagner, Renana Keydar, Omri Abend
Abstract: Effective storytelling relies on a delicate balance between meeting the reader's prior expectations and introducing unexpected developments. In the domain of detective fiction, this tension is known as fair play, which includes the implicit agreement between the writer and the reader as to the range of possible resolutions the mystery story may have. In this work, we present a probabilistic framework for detective fiction that allows us to define desired qualities. Using this framework, we formally define fair play and design appropriate metrics for it. Stemming from these definitions is an inherent tension between the coherence of the story, which measures how much it ``makes sense'', and the surprise it induces. We validate the framework by applying it to LLM-generated detective stories. This domain is appealing since we have an abundance of data, we can sample from the distribution generating the story, and the story-writing capabilities of LLMs are interesting in their own right. Results show that while LLM-generated stories may be unpredictable, they generally fail to balance the trade-off between surprise and fair play, which greatly contributes to their poor quality.
Authors: Nicol\`o Brunello, Davide Rigamonti, Andrea Sassella, Vincenzo Scotti, Mark James Carman
Abstract: The reasoning capabilities of Large Language Models (LLMs) have increased greatly over the last few years, as have their size and complexity. Nonetheless, the use of LLMs in production remains challenging due to their unpredictable nature and discrepancies that can exist between their desired behavior and their actual model output. In this paper, we introduce a new tool, InTraVisTo (Inside Transformer Visualisation Tool), designed to enable researchers to investigate and trace the computational process that generates each token in a Transformer-based LLM. InTraVisTo provides a visualization of both the internal state of the Transformer model (by decoding token embeddings at each layer of the model) and the information flow between the various components across the different layers of the model (using a Sankey diagram). With InTraVisTo, we aim to help researchers and practitioners better understand the computations being performed within the Transformer model and thus to shed some light on internal patterns and reasoning processes employed by LLMs.
Authors: Maciej Jalocha, Johan Hausted Schmidt, William Michelseen
Abstract: The field of cybersecurity NER lacks standardized labels, making it challenging to combine datasets. We investigate label unification across four cybersecurity datasets to increase data resource usability. We perform a coarse-grained label unification and conduct pairwise cross-dataset evaluations using BiLSTM models. Qualitative analysis of predictions reveals errors, limitations, and dataset differences. To address unification limitations, we propose alternative architectures including a multihead model and a graph-based transfer model. Results show that models trained on unified datasets generalize poorly across datasets. The multihead model with weight sharing provides only marginal improvements over unified training, while our graph-based transfer model built on BERT-base-NER shows no significant performance gains compared BERT-base-NER.
Authors: Carlos Mena, Pol Serra, Jacobo Romero, Abir Messaoudi, Jose Giraldo, Carme Armentano-Oller, Rodolfo Zevallos, Ivan Meza, Javier Hernando
Abstract: Code-switching (CS), the alternating use of two or more languages, challenges automatic speech recognition (ASR) due to scarce training data and linguistic similarities. The lack of dedicated CS datasets limits ASR performance, as most models rely on monolingual or mixed-language corpora that fail to reflect real-world CS patterns. This issue is critical in multilingual societies where CS occurs in informal and formal settings. A key example is Catalan-Spanish CS, widely used in media and parliamentary speeches. In this work, we improve ASR for Catalan-Spanish CS by exploring three strategies: (1) generating synthetic CS data, (2) concatenating monolingual audio, and (3) leveraging real CS data with language tokens. We extract CS data from Catalan speech corpora and fine-tune OpenAI's Whisper models, making them available on Hugging Face. Results show that combining a modest amount of synthetic CS data with the dominant language token yields the best transcription performance.
Authors: Cole Walsh, Rodica Ivan, Muhammad Zafar Iqbal, Colleen Robb
Abstract: Academic programs are increasingly recognizing the importance of personal and professional skills and their critical role alongside technical expertise in preparing students for future success in diverse career paths. With this growing demand comes the need for scalable systems to measure, evaluate, and develop these skills. Situational Judgment Tests (SJTs) offer one potential avenue for measuring these skills in a standardized and reliable way, but open-response SJTs have traditionally relied on trained human raters for evaluation, presenting operational challenges to delivering SJTs at scale. Past attempts at developing NLP-based scoring systems for SJTs have fallen short due to issues with construct validity of these systems. In this article, we explore a novel approach to extracting construct-relevant features from SJT responses using large language models (LLMs). We use the Casper SJT to demonstrate the efficacy of this approach. This study sets the foundation for future developments in automated scoring for personal and professional skills.
Authors: Matous Volf (DELTA High school of computer science and economics, Pardubice, Czechia), Jakub Simko (Kempelen Institute of Intelligent Technologies, Bratislava, Slovakia)
Abstract: This paper addresses the challenge of automatically classifying text according to political leaning and politicalness using transformer models. We compose a comprehensive overview of existing datasets and models for these tasks, finding that current approaches create siloed solutions that perform poorly on out-of-distribution texts. To address this limitation, we compile a diverse dataset by combining 12 datasets for political leaning classification and creating a new dataset for politicalness by extending 18 existing datasets with the appropriate label. Through extensive benchmarking with leave-one-in and leave-one-out methodologies, we evaluate the performance of existing models and train new ones with enhanced generalization capabilities.
Authors: Kobi Hackenburg, Ben M. Tappin, Luke Hewitt, Ed Saunders, Sid Black, Hause Lin, Catherine Fist, Helen Margetts, David G. Rand, Christopher Summerfield
Abstract: There are widespread fears that conversational AI could soon exert unprecedented influence over human beliefs. Here, in three large-scale experiments (N=76,977), we deployed 19 LLMs-including some post-trained explicitly for persuasion-to evaluate their persuasiveness on 707 political issues. We then checked the factual accuracy of 466,769 resulting LLM claims. Contrary to popular concerns, we show that the persuasive power of current and near-future AI is likely to stem more from post-training and prompting methods-which boosted persuasiveness by as much as 51% and 27% respectively-than from personalization or increasing model scale. We further show that these methods increased persuasion by exploiting LLMs' unique ability to rapidly access and strategically deploy information and that, strikingly, where they increased AI persuasiveness they also systematically decreased factual accuracy.
Authors: Jan Trienes, Anastasiia Derzhanskaia, Roland Schwarzkopf, Markus M\"uhling, J\"org Schl\"otterer, Christin Seifert
Abstract: We present Marcel, a lightweight and open-source conversational agent designed to support prospective students with admission-related inquiries. The system aims to provide fast and personalized responses, while reducing workload of university staff. We employ retrieval-augmented generation to ground answers in university resources and to provide users with verifiable, contextually relevant information. To improve retrieval quality, we introduce an FAQ retriever that maps user questions to knowledge-base entries, allowing administrators to steer retrieval, and improving over standard dense/hybrid retrieval strategies. The system is engineered for easy deployment in resource-constrained academic settings. We detail the system architecture, provide a technical evaluation of its components, and report insights from a real-world deployment.
Authors: Bianca Raimondi, Maurizio Gabbrielli
Abstract: Large Language Models (LLMs) have become essential in many Natural Language Processing (NLP) tasks, leveraging extensive pre-training and fine-tuning to achieve high accuracy. However, like humans, LLMs exhibit biases, particularly positional biases such as primacy and recency effects, which can influence the accuracy of the answers. The primacy effect-where items presented first are more likely to be remembered or selected-plays a key role in Multiple Choice Question Answering (MCQA), where the order of answer options can affect prediction outcomes. This study focuses on primacy bias in fine-tuned LLMs: We first show that fine-tuning amplifies this bias, probably due to exposure to human-like patterns. Hence, we strategically leverage this effect by reordering response options based on semantic similarity to the query, without requiring knowledge of the correct answer. Our experimental results show that this approach significantly improves performance in MCQA. More generally, our findings underscore the dual nature of biases as both challenges and opportunities, offering insights for bias-aware model design and NLP applications.
Authors: Bhishma Dedhia, Yuval Kansal, Niraj K. Jha
Abstract: Language models traditionally used for cross-domain generalization have recently demonstrated task-specific reasoning. However, their top-down training approach on general corpora is insufficient for acquiring abstractions needed for deep domain expertise. This may require a bottom-up approach that acquires expertise by learning to compose simple domain concepts into more complex ones. A knowledge graph (KG) provides this compositional structure, where domain primitives are represented as head-relation-tail edges and their paths encode higher-level concepts. We present a task generation pipeline that synthesizes tasks directly from KG primitives, enabling models to acquire and compose them for reasoning. We fine-tune language models on the resultant KG-grounded curriculum to demonstrate domain-specific superintelligence. While broadly applicable, we validate our approach in medicine, where reliable KGs exist. Using a medical KG, we curate 24,000 reasoning tasks paired with thinking traces derived from diverse medical primitives. We fine-tune the QwQ-32B model on this curriculum to obtain QwQ-Med-3 that takes a step towards medical superintelligence. We also introduce ICD-Bench, an evaluation suite to quantify reasoning abilities across 15 medical domains. Our experiments demonstrate that QwQ-Med-3 significantly outperforms state-of-the-art reasoning models on ICD-Bench categories. Further analysis reveals that QwQ-Med-3 utilizes acquired primitives to widen the performance gap on the hardest tasks of ICD-Bench. Finally, evaluation on medical question-answer benchmarks shows that QwQ-Med-3 transfers acquired expertise to enhance the base model's performance. While the industry's approach to artificial general intelligence (AGI) emphasizes broad expertise, we envision a future in which AGI emerges from the composable interaction of efficient domain-specific superintelligent agents.
Authors: Lilit Grigoryan, Nikolay Karpov, Enas Albasiri, Vitaly Lavrukhin, Boris Ginsburg
Abstract: Despite Arabic being one of the most widely spoken languages, the development of Arabic Automatic Speech Recognition (ASR) systems faces significant challenges due to the language's complexity, and only a limited number of public Arabic ASR models exist. While much of the focus has been on Modern Standard Arabic (MSA), there is considerably less attention given to the variations within the language. This paper introduces a universal methodology for Arabic speech and text processing designed to address unique challenges of the language. Using this methodology, we train two novel models based on the FastConformer architecture: one designed specifically for MSA and the other, the first unified public model for both MSA and Classical Arabic (CA). The MSA model sets a new benchmark with state-of-the-art (SOTA) performance on related datasets, while the unified model achieves SOTA accuracy with diacritics for CA while maintaining strong performance for MSA. To promote reproducibility, we open-source the models and their training recipes.
Authors: Haoyu He, Haozheng Luo, Yan Chen, Qi R. Wang
Abstract: We introduce RHYTHM (Reasoning with Hierarchical Temporal Tokenization for Human Mobility), a framework that leverages large language models (LLMs) as spatio-temporal predictors and trajectory reasoners. RHYTHM partitions trajectories into daily segments encoded as discrete tokens with hierarchical attention, capturing both daily and weekly dependencies while substantially reducing the sequence length. Token representations are enriched with pre-computed prompt embeddings via a frozen LLM, enhancing the model's ability to capture interdependencies without extensive computational overhead. By freezing the LLM backbone, RHYTHM achieves significant computational efficiency. Evaluation on three real-world datasets demonstrates a 2.4% improvement in accuracy, 5.0% increase on weekends, and 24.6% reduction in training time compared to state-of-the-art methods.
Authors: Jianfei Li, Kevin Kam Fung Yuen
Abstract: This study proposes the Cognitive Pairwise Comparison Classification Model Selection (CPC-CMS) framework for document-level sentiment analysis. The CPC, based on expert knowledge judgment, is used to calculate the weights of evaluation criteria, including accuracy, precision, recall, F1-score, specificity, Matthews Correlation Coefficient (MCC), Cohen's Kappa (Kappa), and efficiency. Naive Bayes, Linear Support Vector Classification (LSVC), Random Forest, Logistic Regression, Extreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), and A Lite Bidirectional Encoder Representations from Transformers (ALBERT) are chosen as classification baseline models. A weighted decision matrix consisting of classification evaluation scores with respect to criteria weights, is formed to select the best classification model for a classification problem. Three open datasets of social media are used to demonstrate the feasibility of the proposed CPC-CMS. Based on our simulation, for evaluation results excluding the time factor, ALBERT is the best for the three datasets; if time consumption is included, no single model always performs better than the other models. The CPC-CMS can be applied to the other classification applications in different areas.
Authors: Israt Jahan, Md Tahmid Rahman Laskar, Chun Peng, Jimmy Huang
Abstract: This paper presents a comprehensive evaluation of cost-efficient Large Language Models (LLMs) for diverse biomedical tasks spanning both text and image modalities. We evaluated a range of closed-source and open-source LLMs on tasks such as biomedical text classification and generation, question answering, and multimodal image processing. Our experimental findings indicate that there is no single LLM that can consistently outperform others across all tasks. Instead, different LLMs excel in different tasks. While some closed-source LLMs demonstrate strong performance on specific tasks, their open-source counterparts achieve comparable results (sometimes even better), with additional benefits like faster inference and enhanced privacy. Our experimental results offer valuable insights for selecting models that are optimally suited for specific biomedical applications.
Authors: Lautaro Estienne, Gabriel Ben Zenou, Nona Naderi, Jackie Cheung, Pablo Piantanida
Abstract: As AI systems take on collaborative roles, they must reason about shared goals and beliefs-not just generate fluent language. The Rational Speech Act (RSA) framework offers a principled approach to pragmatic reasoning, but existing extensions face challenges in scaling to multi-turn, collaborative scenarios. In this paper, we introduce Collaborative Rational Speech Act (CRSA), an information-theoretic (IT) extension of RSA that models multi-turn dialog by optimizing a gain function adapted from rate-distortion theory. This gain is an extension of the gain model that is maximized in the original RSA model but takes into account the scenario in which both agents in a conversation have private information and produce utterances conditioned on the dialog. We demonstrate the effectiveness of CRSA on referential games and template-based doctor-patient dialogs in the medical domain. Empirical results show that CRSA yields more consistent, interpretable, and collaborative behavior than existing baselines-paving the way for more pragmatic and socially aware language agents.
Authors: Garapati Keerthana, Manik Gupta
Abstract: Progress notes are among the most clinically meaningful artifacts in an Electronic Health Record (EHR), offering temporally grounded insights into a patient's evolving condition, treatments, and care decisions. Despite their importance, they are severely underrepresented in large-scale EHR datasets. For instance, in the widely used Medical Information Mart for Intensive Care III (MIMIC-III) dataset, only about $8.56\%$ of hospital visits include progress notes, leaving gaps in longitudinal patient narratives. In contrast, the dataset contains a diverse array of other note types, each capturing different aspects of care. We present DENSE (Documenting Evolving Progress Notes from Scattered Evidence), a system designed to align with clinical documentation workflows by simulating how physicians reference past encounters while drafting progress notes. The system introduces a fine-grained note categorization and a temporal alignment mechanism that organizes heterogeneous notes across visits into structured, chronological inputs. At its core, DENSE leverages a clinically informed retrieval strategy to identify temporally and semantically relevant content from both current and prior visits. This retrieved evidence is used to prompt a large language model (LLM) to generate clinically coherent and temporally aware progress notes. We evaluate DENSE on a curated cohort of patients with multiple visits and complete progress note documentation. The generated notes demonstrate strong longitudinal fidelity, achieving a temporal alignment ratio of $1.089$, surpassing the continuity observed in original notes. By restoring narrative coherence across fragmented documentation, our system supports improved downstream tasks such as summarization, predictive modeling, and clinical decision support, offering a scalable solution for LLM-driven note synthesis in real-world healthcare settings.
Authors: Brian Ondov, William Xia, Kush Attal, Ishita Unde, Jerry He, Hoa Dang, Ian Soboroff, Dina Demner-Fushman
Abstract: Objective: Recent advances in language models have shown potential to adapt professional-facing biomedical literature to plain language, making it accessible to patients and caregivers. However, their unpredictability, combined with the high potential for harm in this domain, means rigorous evaluation is necessary. Our goals with this track were to stimulate research and to provide high-quality evaluation of the most promising systems. Methods: We hosted the Plain Language Adaptation of Biomedical Abstracts (PLABA) track at the 2023 and 2024 Text Retrieval Conferences. Tasks included complete, sentence-level, rewriting of abstracts (Task 1) as well as identifying and replacing difficult terms (Task 2). For automatic evaluation of Task 1, we developed a four-fold set of professionally-written references. Submissions for both Tasks 1 and 2 were provided extensive manual evaluation from biomedical experts. Results: Twelve teams spanning twelve countries participated in the track, with models from multilayer perceptrons to large pretrained transformers. In manual judgments of Task 1, top-performing models rivaled human levels of factual accuracy and completeness, but not simplicity or brevity. Automatic, reference-based metrics generally did not correlate well with manual judgments. In Task 2, systems struggled with identifying difficult terms and classifying how to replace them. When generating replacements, however, LLM-based systems did well in manually judged accuracy, completeness, and simplicity, though not in brevity. Conclusion: The PLABA track showed promise for using Large Language Models to adapt biomedical literature for the general public, while also highlighting their deficiencies and the need for improved automatic benchmarking tools.
Authors: Zeqian Chen
Abstract: The introduction of the transformer architecture in 2017 (cf.\cite{VSP2017}) marked the most striking advancement in natural language processing. The transformer is a model architecture relying entirely on an attention mechanism to draw global dependencies between input and output. However, we believe there is a gap in our theoretical understanding of what the transformer is, and why it works physically. In this paper, from a physical perspective on modern chips, we construct physical models in the Fock space over the Hilbert space of tokens realizing large language models based on a transformer architecture as open quantum systems. Our physical models underlie the transformer architecture for large language models.
Authors: Binbin Ji, Siddharth Agrawal, Qiance Tang, Yvonne Wu
Abstract: This study investigates the spatial reasoning capabilities of vision-language models (VLMs) through Chain-of-Thought (CoT) prompting and reinforcement learning. We begin by evaluating the impact of different prompting strategies and find that simple CoT formats, where the model generates a reasoning step before the answer, not only fail to help, but can even harm the model's original performance. In contrast, structured multi-stage prompting based on scene graphs (SceneGraph CoT) significantly improves spatial reasoning accuracy. Furthermore, to improve spatial reasoning ability, we fine-tune models using Group Relative Policy Optimization (GRPO) on the SAT dataset and evaluate their performance on CVBench. Compared to supervised fine-tuning (SFT), GRPO achieves higher accuracy on Pass@1 evaluations and demonstrates superior robustness under out-of-distribution (OOD) conditions. In particular, we find that SFT overfits to surface-level linguistic patterns and may degrade performance when test-time phrasing changes (e.g., from "closer to" to "farther from"). GRPO, on the other hand, generalizes more reliably and maintains stable performance under such shifts. Our findings provide insights into how reinforcement learning and structured prompting improve the spatial reasoning capabilities and generalization behavior of modern VLMs. All code is open source at: https://github.com/Yvonne511/spatial-vlm-investigator
Authors: Qingyun Sun, Jiaqi Yuan, Shan He, Xiao Guan, Haonan Yuan, Xingcheng Fu, Jianxin Li, Philip S. Yu
Abstract: Graph Retrieval-Augmented Generation has emerged as a powerful paradigm for grounding large language models with external structured knowledge. However, existing Graph RAG methods struggle with temporal reasoning, due to their inability to model the evolving structure and order of real-world events. In this work, we introduce DyG-RAG, a novel event-centric dynamic graph retrieval-augmented generation framework designed to capture and reason over temporal knowledge embedded in unstructured text. To eliminate temporal ambiguity in traditional retrieval units, DyG-RAG proposes Dynamic Event Units (DEUs) that explicitly encode both semantic content and precise temporal anchors, enabling accurate and interpretable time-aware retrieval. To capture temporal and causal dependencies across events, DyG-RAG constructs an event graph by linking DEUs that share entities and occur close in time, supporting efficient and meaningful multi-hop reasoning. To ensure temporally consistent generation, DyG-RAG introduces an event timeline retrieval pipeline that retrieves event sequences via time-aware traversal, and proposes a Time Chain-of-Thought strategy for temporally grounded answer generation. This unified pipeline enables DyG-RAG to retrieve coherent, temporally ordered event sequences and to answer complex, time-sensitive queries that standard RAG systems cannot resolve. Extensive experiments on temporal QA benchmarks demonstrate that DyG-RAG significantly improves the accuracy and recall of three typical types of temporal reasoning questions, paving the way for more faithful and temporal-aware generation. DyG-RAG is available at https://github.com/RingBDStack/DyG-RAG.
Authors: Eduardo C. Garrido-Merch\'an, Cristina Puente
Abstract: The development of large language models (LLMs) has successfully transformed knowledge-based systems such as open domain question nswering, which can automatically produce vast amounts of seemingly coherent information. Yet, those models have several disadvantages like hallucinations or confident generation of incorrect or unverifiable facts. In this paper, we introduce a new approach to the development of expert systems using LLMs in a controlled and transparent way. By limiting the domain and employing a well-structured prompt-based extraction approach, we produce a symbolic representation of knowledge in Prolog, which can be validated and corrected by human experts. This approach also guarantees interpretability, scalability and reliability of the developed expert systems. Via quantitative and qualitative experiments with Claude Sonnet 3.7 and GPT-4.1, we show strong adherence to facts and semantic coherence on our generated knowledge bases. We present a transparent hybrid solution that combines the recall capacity of LLMs with the precision of symbolic systems, thereby laying the foundation for dependable AI applications in sensitive domains.
Authors: Kaiyuan Tang, Kuangshi Ai, Jun Han, Chaoli Wang
Abstract: Advancements in volume visualization (VolVis) focus on extracting insights from 3D volumetric data by generating visually compelling renderings that reveal complex internal structures. Existing VolVis approaches have explored non-photorealistic rendering techniques to enhance the clarity, expressiveness, and informativeness of visual communication. While effective, these methods often rely on complex predefined rules and are limited to transferring a single style, restricting their flexibility. To overcome these limitations, we advocate the representation of VolVis scenes using differentiable Gaussian primitives combined with pretrained large models to enable arbitrary style transfer and real-time rendering. However, conventional 3D Gaussian primitives tightly couple geometry and appearance, leading to suboptimal stylization results. To address this, we introduce TexGS-VolVis, a textured Gaussian splatting framework for VolVis. TexGS-VolVis employs 2D Gaussian primitives, extending each Gaussian with additional texture and shading attributes, resulting in higher-quality, geometry-consistent stylization and enhanced lighting control during inference. Despite these improvements, achieving flexible and controllable scene editing remains challenging. To further enhance stylization, we develop image- and text-driven non-photorealistic scene editing tailored for TexGS-VolVis and 2D-lift-3D segmentation to enable partial editing with fine-grained control. We evaluate TexGS-VolVis both qualitatively and quantitatively across various volume rendering scenes, demonstrating its superiority over existing methods in terms of efficiency, visual quality, and editing flexibility.
Authors: Yanan Wang, Julio Vizcarra, Zhi Li, Hao Niu, Mori Kurokawa
Abstract: Despite recent progress in video large language models (VideoLLMs), a key open challenge remains: how to equip models with chain-of-thought (CoT) reasoning abilities grounded in fine-grained object-level video understanding. Existing instruction-tuned models, such as the Qwen and LLaVA series, are trained on high-level video-text pairs, often lacking structured annotations necessary for compositional, step-by-step reasoning. We propose CoTasks: Chain-of-Thought based Video Instruction Tuning Tasks, a new framework that decomposes complex video questions of existing datasets (e.g., NeXT-QA, STAR) into four entity-level foundational tasks: frame localization, entity tracking, spatial and temporal relation extraction. By embedding these intermediate CoT-style reasoning steps into the input, CoTasks enables models to explicitly perform object-centric spatiotemporal reasoning. Experiments on the NeXT-QA benchmark show that CoTasks significantly enhance inference performance: LLaVA-video-7B improves by +3.3 points in average GPT-4 evaluation score, and Qwen2.5-VL-3B gains +17.4, with large boosts in causal (+14.6), temporal (+10.9), and descriptive (+48.1) subcategories. These results demonstrate the effectiveness of CoTasks as a structured CoT-style supervision framework for improving compositional video reasoning.
Authors: Ye Tian, Xiaoyuan Ren, Zihao Wang, Onat Gungor, Xiaofan Yu, Tajana Rosing
Abstract: Rich and context-aware activity logs facilitate user behavior analysis and health monitoring, making them a key research focus in ubiquitous computing. The remarkable semantic understanding and generation capabilities of Large Language Models (LLMs) have recently created new opportunities for activity log generation. However, existing methods continue to exhibit notable limitations in terms of accuracy, efficiency, and semantic richness. To address these challenges, we propose DailyLLM. To the best of our knowledge, this is the first log generation and summarization system that comprehensively integrates contextual activity information across four dimensions: location, motion, environment, and physiology, using only sensors commonly available on smartphones and smartwatches. To achieve this, DailyLLM introduces a lightweight LLM-based framework that integrates structured prompting with efficient feature extraction to enable high-level activity understanding. Extensive experiments demonstrate that DailyLLM outperforms state-of-the-art (SOTA) log generation methods and can be efficiently deployed on personal computers and Raspberry Pi. Utilizing only a 1.5B-parameter LLM model, DailyLLM achieves a 17% improvement in log generation BERTScore precision compared to the 70B-parameter SOTA baseline, while delivering nearly 10x faster inference speed.
Authors: Pu Jian, Donglei Yu, Wen Yang, Shuo Ren, Jiajun Zhang
Abstract: In visual question answering (VQA) context, users often pose ambiguous questions to visual language models (VLMs) due to varying expression habits. Existing research addresses such ambiguities primarily by rephrasing questions. These approaches neglect the inherently interactive nature of user interactions with VLMs, where ambiguities can be clarified through user feedback. However, research on interactive clarification faces two major challenges: (1) Benchmarks are absent to assess VLMs' capacity for resolving ambiguities through interaction; (2) VLMs are trained to prefer answering rather than asking, preventing them from seeking clarification. To overcome these challenges, we introduce \textbf{ClearVQA} benchmark, which targets three common categories of ambiguity in VQA context, and encompasses various VQA scenarios.
Authors: Shad Nygren, Pinar Avci, Andre Daniels, Reza Rassol, Afshin Beheshti, Diego Galeano
Abstract: Drug side effects are a major global health concern, necessitating advanced methods for their accurate detection and analysis. While Large Language Models (LLMs) offer promising conversational interfaces, their inherent limitations, including reliance on black-box training data, susceptibility to hallucinations, and lack of domain-specific knowledge, hinder their reliability in specialized fields like pharmacovigilance. To address this gap, we propose two architectures: Retrieval-Augmented Generation (RAG) and GraphRAG, which integrate comprehensive drug side effect knowledge into a Llama 3 8B language model. Through extensive evaluations on 19,520 drug side effect associations (covering 976 drugs and 3,851 side effect terms), our results demonstrate that GraphRAG achieves near-perfect accuracy in drug side effect retrieval. This framework offers a highly accurate and scalable solution, signifying a significant advancement in leveraging LLMs for critical pharmacovigilance applications.
Authors: Aleksandr Gashkov, Aleksandr Perevalov, Maria Eltsova, Andreas Both
Abstract: Nowadays, the importance of software with natural-language user interfaces cannot be underestimated. In particular, in Question Answering (QA) systems, generating a SPARQL query for a given natural-language question (often named Query Building) from the information retrieved from the same question is the central task of QA systems working over Knowledge Graphs (KGQA). Due to the rise of Large Language Models (LLMs), they are considered a well-suited method to increase the quality of the question-answering functionality, as there is still a lot of room for improvement, aiming for enhanced quality and trustworthiness. However, LLMs are trained on web data, where researchers have no control over whether the benchmark or the knowledge graph was already included in the training data. In this paper, we introduce a novel method that evaluates the quality of LLMs by generating a SPARQL query from a natural-language question under various conditions: (1) zero-shot SPARQL generation, (2) with knowledge injection, and (3) with "anonymized" knowledge injection. This enables us, for the first time, to estimate the influence of the training data on the QA quality improved by LLMs. Ultimately, this will help to identify how portable a method is or whether good results might mostly be achieved because a benchmark was already included in the training data (cf. LLM memorization). The developed method is portable, robust, and supports any knowledge graph; therefore, it could be easily applied to any KGQA or LLM, s.t., generating consistent insights into the actual LLM capabilities is possible.
Authors: Sichang "Steven" He, Ramesh Govindan, Harsha V. Madhyastha
Abstract: Increasingly, web content is automatically generated by large language models (LLMs) with little human input. We call this "LLM-dominant" content. Since LLMs plagiarize and hallucinate, LLM-dominant content can be unreliable and unethical. Yet, websites rarely disclose such content, and human readers struggle to distinguish it. Thus, we must develop reliable detectors for LLM-dominant content. However, state-of-the-art LLM detectors are insufficient, because they perform well mainly on clean, prose-like text, while web content has complex markup and diverse genres. We propose a highly reliable, scalable pipeline that classifies entire websites. Instead of naively classifying text extracted from each page, we classify each site based on an LLM text detector's outputs of multiple prose-like pages. We train and evaluate our detector by collecting 2 distinct ground truth datasets totaling 120 sites, and obtain 100% accuracies testing across them. In the wild, we detect a sizable portion of sites as LLM-dominant among 10k sites in search engine results and 10k in Common Crawl archives. We find LLM-dominant sites are growing in prevalence and rank highly in search results, raising questions about their impact on end users and the overall Web ecosystem.
Authors: Pawe{\l} Budzianowski, Wesley Maa, Matthew Freed, Jingxiang Mo, Winston Hsiao, Aaron Xie, Tomasz M{\l}oduchowski, Viraj Tipnis, Benjamin Bolte
Abstract: Vision-Language Models (VLMs) have emerged as a promising approach to address the data scarcity challenge in robotics, enabling the development of generalizable visuomotor control policies. While models like OpenVLA showcase the potential of this paradigm, deploying large-scale VLMs on resource-constrained mobile manipulation systems remains a significant hurdle. This paper introduces Edge VLA (EVLA), a novel approach designed to significantly enhance the inference speed of Vision-Language-Action (VLA) models. EVLA maintains the representational power of these models while enabling real-time performance on edge devices. We achieve this through two key innovations: 1) Eliminating the autoregressive requirement for end-effector position prediction, leading to a 7x speedup in inference, and 2) Leveraging the efficiency of Small Language Models (SLMs), demonstrating comparable training performance to larger models with significantly reduced computational demands. Our early results demonstrate that EVLA achieves comparable training characteristics to OpenVLA while offering substantial gains in inference speed and memory efficiency. We release our model checkpoints and training \href{https://github.com/kscalelabs/evla }{codebase} to foster further research.
Authors: Maksim Kuprashevich, Grigorii Alekseenko, Irina Tolstykh, Georgii Fedorov, Bulat Suleimanov, Vladimir Dokholyan, Aleksandr Gordeev
Abstract: Recent advances in generative modeling enable image editing assistants that follow natural language instructions without additional user input. Their supervised training requires millions of triplets: original image, instruction, edited image. Yet mining pixel-accurate examples is hard. Each edit must affect only prompt-specified regions, preserve stylistic coherence, respect physical plausibility, and retain visual appeal. The lack of robust automated edit-quality metrics hinders reliable automation at scale. We present an automated, modular pipeline that mines high-fidelity triplets across domains, resolutions, instruction complexities, and styles. Built on public generative models and running without human intervention, our system uses a task-tuned Gemini validator to score instruction adherence and aesthetics directly, removing any need for segmentation or grounding models. Inversion and compositional bootstrapping enlarge the mined set by approximately 2.2x, enabling large-scale high-fidelity training data. By automating the most repetitive annotation steps, the approach allows a new scale of training without human labeling effort. To democratize research in this resource-intensive area, we release NHR-Edit: an open dataset of 358k high-quality triplets. In the largest cross-dataset evaluation, it surpasses all public alternatives. We also release Bagel-NHR-Edit, an open-source fine-tuned Bagel model, which achieves state-of-the-art metrics in our experiments.
Authors: Son T. Luu, Khoi Trong Hoang, Tuong Quang Pham, Kiet Van Nguyen, Ngan Luu-Thuy Nguyen
Abstract: Machine reading comprehension has been an interesting and challenging task in recent years, with the purpose of extracting useful information from texts. To attain the computer ability to understand the reading text and answer relevant information, we introduce ViMMRC 2.0 - an extension of the previous ViMMRC for the task of multiple-choice reading comprehension in Vietnamese Textbooks which contain the reading articles for students from Grade 1 to Grade 12. This dataset has 699 reading passages which are prose and poems, and 5,273 questions. The questions in the new dataset are not fixed with four options as in the previous version. Moreover, the difficulty of questions is increased, which challenges the models to find the correct choice. The computer must understand the whole context of the reading passage, the question, and the content of each choice to extract the right answers. Hence, we propose a multi-stage approach that combines the multi-step attention network (MAN) with the natural language inference (NLI) task to enhance the performance of the reading comprehension model. Then, we compare the proposed methodology with the baseline BERTology models on the new dataset and the ViMMRC 1.0. From the results of the error analysis, we found that the challenge of the reading comprehension models is understanding the implicit context in texts and linking them together in order to find the correct answers. Finally, we hope our new dataset will motivate further research to enhance the ability of computers to understand the Vietnamese language.
Authors: Elisa Sanchez-Bayona, Rodrigo Agerri
Abstract: Metaphors, although occasionally unperceived, are ubiquitous in our everyday language. Thus, it is crucial for Language Models to be able to grasp the underlying meaning of this kind of figurative language. In this work, we present Meta4XNLI, a novel parallel dataset for the tasks of metaphor detection and interpretation that contains metaphor annotations in both Spanish and English. We investigate language models' metaphor identification and understanding abilities through a series of monolingual and cross-lingual experiments by leveraging our proposed corpus. In order to comprehend how these non-literal expressions affect models' performance, we look over the results and perform an error analysis. Additionally, parallel data offers many potential opportunities to investigate metaphor transferability between these languages and the impact of translation on the development of multilingual annotated resources.
Authors: Guillaume Rochette, Mathieu Rochat, Matthew J. Vowels
Abstract: psifx is a plug-and-play multi-modal feature extraction toolkit, aiming to facilitate and democratize the use of state-of-the-art machine learning techniques for human sciences research. It is motivated by a need (a) to automate and standardize data annotation processes that typically require expensive, lengthy, and inconsistent human labour; (b) to develop and distribute open-source community-driven psychology research software; and (c) to enable large-scale access and ease of use for non-expert users. The framework contains an array of tools for tasks such as speaker diarization, closed-caption transcription and translation from audio; body, hand, and facial pose estimation and gaze tracking with multi-person tracking from video; and interactive textual feature extraction supported by large language models. The package has been designed with a modular and task-oriented approach, enabling the community to add or update new tools easily. This combination creates new opportunities for in-depth study of real-time behavioral phenomena in psychological and social science research.
Authors: Barrett Martin Lattimer, Varun Gangal, Ryan McDonald, Yi Yang
Abstract: Recent advancements in state-of-the-art (SOTA) Large Language Model (LLM) agents, especially in multi-turn dialogue tasks, have been primarily driven by supervised fine-tuning and high-quality human feedback. However, as base LLM models continue to improve, acquiring meaningful human feedback has become increasingly challenging and costly. In certain domains, base LLM agents may eventually exceed human capabilities, making traditional feedback-driven methods impractical. In this paper, we introduce a novel self-improvement paradigm that empowers LLM agents to autonomously enhance their performance without external human feedback. Our method, Juxtaposed Outcomes for Simulation Harvesting (JOSH), is a self-alignment algorithm that leverages a sparse reward simulation environment to extract ideal behaviors and further train the LLM on its own outputs. We present ToolWOZ, a sparse reward tool-calling simulation environment derived from MultiWOZ. We demonstrate that models trained with JOSH, both small and frontier, significantly improve tool-based interactions while preserving general model capabilities across diverse benchmarks. Our code and data are publicly available on GitHub at https://github.com/asappresearch/josh-llm-simulation-training
URLs: https://github.com/asappresearch/josh-llm-simulation-training
Authors: Sumanth Doddapaneni, Mohammed Safi Ur Rahman Khan, Dilip Venkatesh, Raj Dabre, Anoop Kunchukuttan, Mitesh M. Khapra
Abstract: Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English languages. Current methodologies, including automated metrics, human assessments, and LLM-based evaluations, predominantly focus on English, revealing a significant gap in multilingual evaluation frameworks. We introduce the Cross Lingual Auto Evaluation (CIA) Suite, an extensible framework that includes evaluator LLMs (Hercule) and a novel test set (Recon) specifically designed for multilingual evaluation. Our test set features 500 human-annotated instructions spanning various task capabilities along with human judgment scores across six languages. This would enable benchmarking of general-purpose multilingual LLMs and facilitate meta-evaluation of Evaluator LLMs. The proposed model, Hercule, is a cross-lingual evaluation model that addresses the scarcity of reference answers in the target language by learning to assign scores to responses based on easily available reference answers in English. Our experiments demonstrate that Hercule aligns more closely with human judgments compared to proprietary models, demonstrating the effectiveness of such cross-lingual evaluation in low resource scenarios. Further, it is also effective in zero-shot evaluation on unseen languages. This study is the first comprehensive examination of cross-lingual evaluation using LLMs, presenting a scalable and effective approach for multilingual assessment. All code, datasets, and models will be publicly available to enable further research in this important area.
Authors: Jiayu Song, Mahmud Elahi Akhter, Dana Atzil Slonim, Maria Liakata
Abstract: This paper explores whether enhancing temporal reasoning capabilities in Large Language Models (LLMs) can improve the quality of timeline summarisation, the task of summarising long texts containing sequences of events, such as social media threads. We first introduce NarrativeReason, a novel dataset focused on temporal relationships among sequential events within narratives, distinguishing it from existing temporal reasoning datasets that primarily address pair-wise event relationships. Our approach then combines temporal reasoning with timeline summarisation through a knowledge distillation framework, where we first fine-tune a teacher model on temporal reasoning tasks and then distill this knowledge into a student model while simultaneously training it for the task of timeline summarisation. Experimental results demonstrate that our model achieves superior performance on out-of-domain mental health-related timeline summarisation tasks, which involve long social media threads with repetitions of events and a mix of emotions, highlighting the importance and generalisability of leveraging temporal reasoning to improve timeline summarisation.
Authors: Wenlu Fan, Yuqi Zhu, Chenyang Wang, Bin Wang, Wentao Xu
Abstract: Large Language Models (LLMs) demonstrate remarkable capabilities in text generation, yet their emotional consistency and semantic coherence in social media contexts remain insufficiently understood. This study investigates how LLMs handle emotional content and maintain semantic relationships through continuation and response tasks using two open-source models: Gemma and Llama. By analyzing climate change discussions from Twitter and Reddit, we examine emotional transitions, intensity patterns, and semantic similarity between human-authored and LLM-generated content. Our findings reveal that while both models maintain high semantic coherence, they exhibit distinct emotional patterns: Gemma shows a tendency toward negative emotion amplification, particularly anger, while maintaining certain positive emotions like optimism. Llama demonstrates superior emotional preservation across a broader spectrum of affects. Both models systematically generate responses with attenuated emotional intensity compared to human-authored content and show a bias toward positive emotions in response tasks. Additionally, both models maintain strong semantic similarity with original texts, though performance varies between continuation and response tasks. These findings provide insights into LLMs' emotional and semantic processing capabilities, with implications for their deployment in social media contexts and human-AI interaction design.
Authors: Mohita Chowdhury, Yajie Vera He, Jared Joselowitz, Aisling Higham, Ernest Lim
Abstract: Large Language Models (LLMs) have shown impressive potential in clinical question answering (QA), with Retrieval Augmented Generation (RAG) emerging as a leading approach for ensuring the factual accuracy of model responses. However, current automated RAG metrics perform poorly in clinical and conversational use cases. Using clinical human evaluations of responses is expensive, unscalable, and not conducive to the continuous iterative development of RAG systems. To address these challenges, we introduce ASTRID - an Automated and Scalable TRIaD for evaluating clinical QA systems leveraging RAG - consisting of three metrics: Context Relevance (CR), Refusal Accuracy (RA), and Conversational Faithfulness (CF). Our novel evaluation metric, CF, is designed to better capture the faithfulness of a model's response to the knowledge base without penalising conversational elements. To validate our triad, we curate a dataset of over 200 real-world patient questions posed to an LLM-based QA agent during surgical follow-up for cataract surgery - the highest volume operation in the world - augmented with clinician-selected questions for emergency, clinical, and non-clinical out-of-domain scenarios. We demonstrate that CF can predict human ratings of faithfulness better than existing definitions for conversational use cases. Furthermore, we show that evaluation using our triad consisting of CF, RA, and CR exhibits alignment with clinician assessment for inappropriate, harmful, or unhelpful responses. Finally, using nine different LLMs, we demonstrate that the three metrics can closely agree with human evaluations, highlighting the potential of these metrics for use in LLM-driven automated evaluation pipelines. We also publish the prompts and datasets for these experiments, providing valuable resources for further research and development.
Authors: Naitian Zhou, David Bamman, Isaac L. Bleaman
Abstract: The field of cultural NLP has recently experienced rapid growth, driven by a pressing need to ensure that language technologies are effective and safe across a pluralistic user base. This work has largely progressed without a shared conception of culture, instead choosing to rely on a wide array of cultural proxies. However, this leads to a number of recurring limitations: coarse national boundaries fail to capture nuanced differences that lay within them, limited coverage restricts datasets to only a subset of usually highly-represented cultures, and a lack of dynamicity results in static cultural benchmarks that do not change as culture evolves. In this position paper, we argue that these methodological limitations are symptomatic of a theoretical gap. We draw on a well-developed theory of culture from sociocultural linguistics to fill this gap by 1) demonstrating in a case study how it can clarify methodological constraints and affordances, 2) offering theoretically-motivated paths forward to achieving cultural competence, and 3) arguing that localization is a more useful framing for the goals of much current work in cultural NLP.
Authors: Julia Mendelsohn, Ceren Budak
Abstract: Metaphor, discussing one concept in terms of another, is abundant in politics and can shape how people understand important issues. We develop a computational approach to measure metaphorical language, focusing on immigration discourse on social media. Grounded in qualitative social science research, we identify seven concepts evoked in immigration discourse (e.g. "water" or "vermin"). We propose and evaluate a novel technique that leverages both word-level and document-level signals to measure metaphor with respect to these concepts. We then study the relationship between metaphor, political ideology, and user engagement in 400K US tweets about immigration. While conservatives tend to use dehumanizing metaphors more than liberals, this effect varies widely across concepts. Moreover, creature-related metaphor is associated with more retweets, especially for liberal authors. Our work highlights the potential for computational methods to complement qualitative approaches in understanding subtle and implicit language in political discourse.
Authors: William Jurayj, Jeffrey Cheng, Benjamin Van Durme
Abstract: Scaling the test-time compute of large language models has demonstrated impressive performance on reasoning benchmarks. However, existing evaluations of test-time scaling make the strong assumption that a reasoning system should always give an answer to any question provided. This overlooks concerns about whether a model is confident in its answer, and whether it is appropriate to always provide a response. To address these concerns, we extract confidence scores during reasoning for thresholding model responses. We find that increasing compute budget at inference time not only helps models answer more questions correctly, but also increases confidence in correct responses. We then extend the current paradigm of zero-risk responses during evaluation by considering settings with non-zero levels of response risk, and suggest a recipe for reporting evaluations under these settings.
Authors: Jie Ouyang, Tingyue Pan, Mingyue Cheng, Ruiran Yan, Yucong Luo, Jiaying Lin, Qi Liu
Abstract: While Retrieval-Augmented Generation (RAG) has emerged as an effective approach for addressing the knowledge outdating problem in Large Language Models (LLMs), it still faces a critical challenge: the prevalence of outdated information in knowledge bases. Current research primarily focuses on incorporating up-to-date information, yet the impact of outdated information coexisting in retrieval sources remains inadequately addressed. To bridge this gap, we introduce HoH, the first benchmark specifically designed to evaluate the impact of outdated information on RAG. Our benchmark leverages token-level diff algorithms combined with LLM pipelines to efficiently create a large-scale QA dataset that accurately captures the evolution of temporal knowledge in real-world facts. Through comprehensive experiments, we reveal that outdated information significantly degrades RAG performance in two critical ways: (1) it substantially reduces response accuracy by distracting models from correct information, and (2) it can mislead models into generating potentially harmful outputs, even when current information is available. Current RAG approaches struggle with both retrieval and generation aspects when handling outdated information. These findings highlight the urgent need for innovative solutions to address the temporal challenges in RAG. Our code and data are available at: https://github.com/0russwest0/HoH.
Authors: Jaap Jumelet, Leonie Weissweiler, Joakim Nivre, Arianna Bisazza
Abstract: We introduce MultiBLiMP 1.0, a massively multilingual benchmark of linguistic minimal pairs, covering 101 languages and 2 types of subject-verb agreement, containing more than 128,000 minimal pairs. Our minimal pairs are created using a fully automated pipeline, leveraging the large-scale linguistic resources of Universal Dependencies and UniMorph. MultiBLiMP 1.0 evaluates abilities of LLMs at an unprecedented multilingual scale, and highlights the shortcomings of the current state-of-the-art in modelling low-resource languages
Authors: Tong Chen, Faeze Brahman, Jiacheng Liu, Niloofar Mireshghallah, Weijia Shi, Pang Wei Koh, Luke Zettlemoyer, Hannaneh Hajishirzi
Abstract: Language models (LMs) can memorize and reproduce segments from their pretraining data verbatim even in non-adversarial settings, raising concerns about copyright, plagiarism, privacy, and creativity. We introduce Paraphrase Preference Optimization (ParaPO), a post-training method that fine-tunes LMs to reduce unintentional regurgitation while preserving their overall utility. ParaPO trains LMs to prefer paraphrased versions of memorized segments over the original verbatim content from the pretraining data. To maintain the ability to recall famous quotations when appropriate, we develop a variant of ParaPO that uses system prompts to control regurgitation behavior. In our evaluation on Llama3.1-8B, ParaPO consistently reduces regurgitation across all tested datasets (e.g., reducing the regurgitation metric from 17.3 to 12.9 in creative writing), whereas unlearning methods used in prior work to mitigate regurgitation are less effective outside their targeted unlearned domain (from 17.3 to 16.9). When applied to the instruction-tuned Tulu3-8B model, ParaPO with system prompting successfully preserves famous quotation recall while reducing unintentional regurgitation (from 8.7 to 6.3 in creative writing) when prompted not to regurgitate. In contrast, without ParaPO tuning, prompting the model not to regurgitate produces only a marginal reduction (8.7 to 8.4).
Authors: Z. Z. Ren, Zhihong Shao, Junxiao Song, Huajian Xin, Haocheng Wang, Wanjia Zhao, Liyue Zhang, Zhe Fu, Qihao Zhu, Dejian Yang, Z. F. Wu, Zhibin Gou, Shirong Ma, Hongxuan Tang, Yuxuan Liu, Wenjun Gao, Daya Guo, Chong Ruan
Abstract: We introduce DeepSeek-Prover-V2, an open-source large language model designed for formal theorem proving in Lean 4, with initialization data collected through a recursive theorem proving pipeline powered by DeepSeek-V3. The cold-start training procedure begins by prompting DeepSeek-V3 to decompose complex problems into a series of subgoals. The proofs of resolved subgoals are synthesized into a chain-of-thought process, combined with DeepSeek-V3's step-by-step reasoning, to create an initial cold start for reinforcement learning. This process enables us to integrate both informal and formal mathematical reasoning into a unified model. The resulting model, DeepSeek-Prover-V2-671B, achieves state-of-the-art performance in neural theorem proving, reaching 88.9% pass ratio on the MiniF2F-test and solving 49 out of 658 problems from PutnamBench. In addition to standard benchmarks, we introduce ProverBench, a collection of 325 formalized problems, to enrich our evaluation, including 15 selected problems from the recent AIME competitions (years 24-25). Further evaluation on these 15 AIME problems shows that the model successfully solves 6 of them. In comparison, DeepSeek-V3 solves 8 of these problems using majority voting, highlighting that the gap between formal and informal mathematical reasoning in large language models is substantially narrowing.
Authors: Haoyuan Wu, Rui Ming, Jilong Gao, Hangyu Zhao, Xueyi Chen, Yikai Yang, Haisheng Zheng, Zhuolun He, Bei Yu
Abstract: Large language models (LLMs) achieve remarkable performance in code generation tasks. However, a significant performance disparity persists between popular programming languages (e.g., Python, C++) and others. To address this capability gap, we leverage the code translation task to train LLMs, thereby facilitating the transfer of coding proficiency across diverse programming languages. Moreover, we introduce OORL for training, a novel reinforcement learning (RL) framework that integrates on-policy and off-policy strategies. Within OORL, on-policy RL is applied during code translation, guided by a rule-based reward signal derived from unit tests. Complementing this coarse-grained rule-based reward, we propose Group Equivalent Preference Optimization (GEPO), a novel preference optimization method. Specifically, GEPO trains the LLM using intermediate representations (IRs) groups. LLMs can be guided to discern IRs equivalent to the source code from inequivalent ones, while also utilizing signals about the mutual equivalence between IRs within the group. This process allows LLMs to capture nuanced aspects of code functionality. By employing OORL for training with code translation tasks, LLMs improve their recognition of code functionality and their understanding of the relationships between code implemented in different languages. Extensive experiments demonstrate that our OORL for LLMs training with code translation tasks achieves significant performance improvements on code benchmarks across multiple programming languages.
Authors: Michael Sullivan
Abstract: We make the case for language models over logical forms (LFLMs), arguing that such models are more data-efficient than their textual counterparts. To that end, we introduce the Graph-based Formal-Logical Distributional Semantics (GFoLDS) prototype, a pretrained LM over graph representations of logical forms, as a proof-of-concept of LFLMs. Using GFoLDS, we present strong experimental evidence that LFLMs can leverage the built-in, basic linguistic knowledge inherent in such models to immediately begin learning more complex patterns. On downstream tasks, we show that GFoLDS vastly outperforms textual, transformer LMs (BERT) pretrained on the same data, indicating that LFLMs can learn with substantially less data than models over plain text. Furthermore, we show that the performance of this model is likely to scale with additional parameters and pretraining data, suggesting the viability of LFLMs in real-world applications.
Authors: Ramon Ferrer-i-Cancho
Abstract: Here we present a new class of optimality for coding systems. Members of that class are displaced linearly from optimal coding and thus exhibit Zipf's law, namely a power-law distribution of frequency ranks. Within that class, Zipf's law, the size-rank law and the size-probability law form a group-like structure. We identify human languages that are members of the class. All languages showing sufficient agreement with Zipf's law are potential members of the class. In contrast, there are communication systems in other species that cannot be members of that class for exhibiting an exponential distribution instead but dolphins and humpback whales might. We provide a new insight into plots of frequency versus rank in double logarithmic scale. For any system, a straight line in that scale indicates that the lengths of optimal codes under non-singular coding and under uniquely decodable encoding are displaced by a linear function whose slope is the exponent of Zipf's law. For systems under compression and constrained to be uniquely decodable, such a straight line may indicate that the system is coding close to optimality. We provide support for the hypothesis that Zipf's law originates from compression and define testable conditions for the emergence of Zipf's law in compressing systems.
Authors: Nicholas Edwards, Yukyung Lee, Yujun Audrey Mao, Yulu Qin, Sebastian Schuster, Najoung Kim
Abstract: Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of 12 realistic research experiment implementation tasks that aim to investigate research hypotheses that have not previously been implemented. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate nine LLM agents implemented using three different frameworks: aider, Claude Code, and OpenHands. We find that all agents evaluated fail to autonomously implement the majority of the extensions. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 40%. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance.
Authors: Ian R. McKenzie, Oskar J. Hollinsworth, Tom Tseng, Xander Davies, Stephen Casper, Aaron D. Tucker, Robert Kirk, Adam Gleave
Abstract: Frontier AI developers are relying on layers of safeguards to protect against catastrophic misuse of AI systems. Anthropic guards their latest Claude 4 Opus model using one such defense pipeline, and other frontier developers including Google DeepMind and OpenAI pledge to soon deploy similar defenses. However, the security of such pipelines is unclear, with limited prior work evaluating or attacking these pipelines. We address this gap by developing and red-teaming an open-source defense pipeline. First, we find that a novel few-shot-prompted input and output classifier outperforms state-of-the-art open-weight safeguard model ShieldGemma across three attacks and two datasets, reducing the attack success rate (ASR) to 0% on the catastrophic misuse dataset ClearHarm. Second, we introduce a STaged AttaCK (STACK) procedure that achieves 71% ASR on ClearHarm in a black-box attack against the few-shot-prompted classifier pipeline. Finally, we also evaluate STACK in a transfer setting, achieving 33% ASR, providing initial evidence that it is feasible to design attacks with no access to the target pipeline. We conclude by suggesting specific mitigations that developers could use to thwart staged attacks.
Authors: Xiangru Tang, Tianrui Qin, Tianhao Peng, Ziyang Zhou, Daniel Shao, Tingting Du, Xinming Wei, Peng Xia, Fang Wu, He Zhu, Ge Zhang, Jiaheng Liu, Xingyao Wang, Sirui Hong, Chenglin Wu, Hao Cheng, Chi Wang, Wangchunshu Zhou
Abstract: Current AI agents cannot effectively learn from each other's problem-solving experiences or use past successes to guide self-reflection and error correction in new tasks. We introduce Agent KB, a shared knowledge base that captures both high-level problem-solving strategies and detailed execution lessons, enabling knowledge transfer across agent frameworks. Agent KB implements a novel teacher-student dual-phase retrieval mechanism where student agents retrieve workflow-level patterns for strategic guidance while teacher agents identify execution-level patterns for refinement. This hierarchical approach enables agents to break out of limited reasoning pathways by incorporating diverse strategies from external sources. Evaluations on the GAIA benchmark demonstrate substantial performance gains, with Agent KB improving success rates by up to 6.06 percentage points overall under pass@1. For SWE-bench code repair tasks, our system significantly improved resolution rates, with o3-mini achieving an 8.67 percentage point gain (23 percent to 31.67 percent) in pass@1.
Authors: Seokhee Hong, Sunkyoung Kim, Guijin Son, Soyeon Kim, Yeonjung Hong, Jinsik Lee
Abstract: The development of Large Language Models (LLMs) requires robust benchmarks that encompass not only academic domains but also industrial fields to effectively evaluate their applicability in real-world scenarios. In this paper, we introduce two Korean expert-level benchmarks. KMMLU-Redux, reconstructed from the existing KMMLU, consists of questions from the Korean National Technical Qualification exams, with critical errors removed to enhance reliability. KMMLU-Pro is based on Korean National Professional Licensure exams to reflect professional knowledge in Korea. Our experiments demonstrate that these benchmarks comprehensively represent industrial knowledge in Korea. We release our dataset publicly available.
Authors: Lionel Wong, Katherine M. Collins, Lance Ying, Cedegao E. Zhang, Adrian Weller, Tobias Gerstenberg, Timothy O'Donnell, Alexander K. Lew, Jacob D. Andreas, Joshua B. Tenenbaum, Tyler Brooke-Wilson
Abstract: When faced with novel situations, people are able to marshal relevant considerations from a wide range of background knowledge and put these to use in inferences and predictions. What permits us to draw in globally relevant information and reason over it coherently? Here, we explore the hypothesis that people use a combination of distributed and symbolic representations to construct bespoke mental models tailored to novel situations. We propose a computational implementation of this idea -- a ``Model Synthesis Architecture'' (MSA) -- using language models to implement global relevance-based retrieval and model synthesis and probabilistic programs to implement bespoke, coherent world models. We evaluate our MSA as a model of human judgments on a novel reasoning dataset. The dataset -- built around a `Model Olympics` domain of sports vignettes -- tests models' capacity for human-like, open-ended reasoning by requiring (i) judgments about novel causal structures described in language; (ii) drawing on large bodies of background knowledge; and (iii) doing both in light of observations that introduce arbitrary novel variables. Our MSA approach captures human judgments better than language model-only baselines, under both direct and chain-of-thought generations from the LM that supports model synthesis. These results suggest that MSAs can be implemented in a way that mirrors people's ability to deliver locally coherent reasoning over globally relevant variables, offering a path to understanding and replicating human reasoning in open-ended domains.
Authors: Retief Louw, Emma Sharratt, Febe de Wet, Christiaan Jacobs, Annelien Smith, Herman Kamper
Abstract: Developing narrative and comprehension skills in early childhood is critical for later literacy. However, teachers in large preschool classrooms struggle to accurately identify students who require intervention. We present a system for automatically assessing oral narratives of preschool children in Afrikaans and isiXhosa. The system uses automatic speech recognition followed by a machine learning scoring model to predict narrative and comprehension scores. For scoring predicted transcripts, we compare a linear model to a large language model (LLM). The LLM-based system outperforms the linear model in most cases, but the linear system is competitive despite its simplicity. The LLM-based system is comparable to a human expert in flagging children who require intervention. We lay the foundation for automatic oral assessments in classrooms, giving teachers extra capacity to focus on personalised support for children's learning.
Authors: Wan-Cyuan Fan, Yen-Chun Chen, Mengchen Liu, Lu Yuan, Leonid Sigal
Abstract: Recent studies customizing Multimodal Large Language Models (MLLMs) for domain-specific tasks have yielded promising results, especially in the field of scientific chart comprehension. These studies generally utilize visual instruction tuning with specialized datasets to enhance question and answer (QA) accuracy within the chart domain. However, they often neglect the fundamental discrepancy between natural image-caption pre-training data and digital chart image-QA data, particularly in the models' capacity to extract underlying numeric values from charts. This paper tackles this oversight by exploring the training processes necessary to improve MLLMs' comprehension of charts. We present three key findings: (1) Incorporating raw data values in alignment pre-training markedly improves comprehension of chart data. (2) Replacing images with their textual representation randomly during end-to-end fine-tuning transfer the language reasoning capability to chart interpretation skills. (3) Requiring the model to first extract the underlying chart data and then answer the question in the fine-tuning can further improve the accuracy. Consequently, we introduce CHOPINLLM, an MLLM tailored for in-depth chart comprehension. CHOPINLLM effectively interprets various types of charts, including unannotated ones, while maintaining robust reasoning abilities. Furthermore, we establish a new benchmark to evaluate MLLMs' understanding of different chart types across various comprehension levels. Experimental results show that CHOPINLLM exhibits strong performance in understanding both annotated and unannotated charts across a wide range of types.
Authors: Alexander Michael Rombach, Peter Fettke
Abstract: Extracting key information from documents represents a large portion of business workloads and therefore offers a high potential for efficiency improvements and process automation. With recent advances in Deep Learning, a plethora of Deep Learning based approaches for Key Information Extraction have been proposed under the umbrella term Document Understanding that enable the processing of complex business documents. The goal of this systematic literature review is an in-depth analysis of existing approaches in this domain and the identification of opportunities for further research. To this end, 130 approaches published between 2017 and 2024 are analyzed in this study.
Authors: Ebube Alor, Ahmad Abdellatif, SayedHassan Khatoonabadi, Emad Shihab
Abstract: Software engineering (SE) chatbots are increasingly gaining attention for their role in enhancing development processes. At the core of chatbots are Natural Language Understanding platforms (NLUs), which enable them to comprehend user queries but require labeled data for training. However, acquiring such labeled data for SE chatbots is challenging due to the scarcity of high-quality datasets, as training requires specialized vocabulary and phrases not found in typical language datasets. Consequently, developers often resort to manually annotating user queries -- a time-consuming and resource-intensive process. Previous approaches require human intervention to generate rules, called labeling functions (LFs), that categorize queries based on specific patterns. To address this issue, we propose an approach to automatically generate LFs by extracting patterns from labeled user queries. We evaluate our approach on four SE datasets and measure performance improvement from training NLUs on queries labeled by the generated LFs. The generated LFs effectively label data with AUC scores up to 85.3% and NLU performance improvements up to 27.2%. Furthermore, our results show that the number of LFs affects labeling performance. We believe that our approach can save time and resources in labeling users' queries, allowing practitioners to focus on core chatbot functionalities rather than manually labeling queries.
Authors: Ammar Ahmed, Margarida Fresco, Fredrik Forsberg, Hallvard Grotli
Abstract: Web accessibility ensures that individuals with disabilities can access and interact with digital content without barriers, yet a significant majority of most used websites fail to meet accessibility standards. This study evaluates ChatGPT's (GPT-4o) ability to generate and improve web pages in line with Web Content Accessibility Guidelines (WCAG). While ChatGPT can effectively address accessibility issues when prompted, its default code often lacks compliance, reflecting limitations in its training data and prevailing inaccessible web practices. Automated and manual testing revealed strengths in resolving simple issues but challenges with complex tasks, requiring human oversight and additional iterations. Unlike prior studies, we incorporate manual evaluation, dynamic elements, and use the visual reasoning capability of ChatGPT along with the prompts to fix accessibility issues. Providing screenshots alongside prompts enhances the LLM's ability to address accessibility issues by allowing it to analyze surrounding components, such as determining appropriate contrast colors. We found that effective prompt engineering, such as providing concise, structured feedback and incorporating visual aids, significantly enhances ChatGPT's performance. These findings highlight the potential and limitations of large language models for accessible web development, offering practical guidance for developers to create more inclusive websites.
Authors: Raghav Singhal, Zachary Horvitz, Ryan Teehan, Mengye Ren, Zhou Yu, Kathleen McKeown, Rajesh Ranganath
Abstract: Diffusion models produce impressive results in modalities ranging from images and video to protein design and text. However, generating samples with user-specified properties remains a challenge. Recent research proposes fine-tuning models to maximize rewards that capture desired properties, but these methods require expensive training and are prone to mode collapse. In this work, we present Feynman-Kac (FK) steering, an inference-time framework for steering diffusion models with reward functions. FK steering works by sampling a system of multiple interacting diffusion processes, called particles, and resampling particles at intermediate steps based on scores computed using functions called potentials. Potentials are defined using rewards for intermediate states and are selected such that a high value indicates that the particle will yield a high-reward sample. We explore various choices of potentials, intermediate rewards, and samplers. We evaluate FK steering on text-to-image and text diffusion models. For steering text-to-image models with a human preference reward, we find that FK steering a 0.8B parameter model outperforms a 2.6B parameter fine-tuned model on prompt fidelity, with faster sampling and no training. For steering text diffusion models with rewards for text quality and specific text attributes, we find that FK steering generates lower perplexity, more linguistically acceptable outputs and enables gradient-free control of attributes like toxicity. Our results demonstrate that inference-time scaling and steering of diffusion models - even with off-the-shelf rewards - can provide significant sample quality gains and controllability benefits. Code is available at https://github.com/zacharyhorvitz/Fk-Diffusion-Steering .
URLs: https://github.com/zacharyhorvitz/Fk-Diffusion-Steering
Authors: Wannita Takerngsaksiri, Chakkrit Tantithamthavorn, Micheal Fu, Jirat Pasuksmit, Kun Chen, Ming Wu
Abstract: Software engineers spend a significant amount of time reading code during the software development process, especially in the age of large language models (LLMs) that can automatically generate code. However, little is known about the readability of the LLM-generated code and whether it is still important from practitioners' perspectives in this new era. In this paper, we conduct a survey to explore the practitioners' perspectives on code readability in the age of LLMs and investigate the readability of our LLM-based software development agents framework, HULA, by comparing its generated code with human-written code in real-world scenarios. Overall, the findings underscore that (1) readability remains a critical aspect of software development; (2) the readability of our LLM-generated code is comparable to human-written code, fostering the establishment of appropriate trust and driving the broad adoption of our LLM-powered software development platform.
Authors: Haozhe Wang, Long Li, Chao Qu, Fengming Zhu, Weidi Xu, Wei Chu, Fangzhen Lin
Abstract: Recent advances in mathematical problem-solving with language models (LMs) integrate chain-of-thought (CoT) reasoning and code execution to harness their complementary strengths. However, existing hybrid frameworks exhibit a critical limitation: they depend on externally dictated instructions or rigid code-integration templates, lacking metacognitive awareness -- the capacity to dynamically evaluate intrinsic capabilities and autonomously determine when and how to integrate tools. This rigidity motivates our study of autonomous code integration, enabling models to adapt tool-usage strategies as their reasoning abilities evolve during training. While reinforcement learning (RL) shows promise for boosting LLM reasoning at scale (e.g., DeepSeek-R1), we demonstrate its inefficiency in learning autonomous code integration due to inadequate exploration of the vast combinatorial space of CoT-code interleaving patterns. To address this challenge, we propose a novel Expectation-Maximization (EM) framework that synergizes structured exploration (E-step) with off-policy RL optimization (M-step), creating a self-reinforcing cycle between metacognitive tool-use decisions and evolving capabilities. Experiments reveal our method achieves superior results through improved exploration. Notably, our 7B model improves over 11% on MATH500 and 9.4% on AIME without o1-like CoT.
Authors: Yuan Gao, Mattia Piccinini, Korbinian Moller, Amr Alanwar, Johannes Betz
Abstract: Ensuring the safety of autonomous vehicles requires virtual scenario-based testing, which depends on the robust evaluation and generation of safety-critical scenarios. So far, researchers have used scenario-based testing frameworks that rely heavily on handcrafted scenarios as safety metrics. To reduce the effort of human interpretation and overcome the limited scalability of these approaches, we combine Large Language Models (LLMs) with structured scenario parsing and prompt engineering to automatically evaluate and generate safety-critical driving scenarios. We introduce Cartesian and Ego-centric prompt strategies for scenario evaluation, and an adversarial generation module that modifies trajectories of risk-inducing vehicles (ego-attackers) to create critical scenarios. We validate our approach using a 2D simulation framework and multiple pre-trained LLMs. The results show that the evaluation module effectively detects collision scenarios and infers scenario safety. Meanwhile, the new generation module identifies high-risk agents and synthesizes realistic, safety-critical scenarios. We conclude that an LLM equipped with domain-informed prompting techniques can effectively evaluate and generate safety-critical driving scenarios, reducing dependence on handcrafted metrics. We release our open-source code and scenarios at: https://github.com/TUM-AVS/From-Words-to-Collisions.
Authors: Qitao Tan, Jun Liu, Zheng Zhan, Caiwei Ding, Yanzhi Wang, Xiaolong Ma, Jaewoo Lee, Jin Lu, Geng Yuan
Abstract: Large language models (LLMs) excel across various tasks, but standard first-order (FO) fine-tuning demands considerable memory, significantly limiting real-world deployment. Recently, zeroth-order (ZO) optimization stood out as a promising memory-efficient training paradigm, avoiding backward passes and relying solely on forward passes for gradient estimation, making it attractive for resource-constrained scenarios. However, ZO method lags far behind FO method in both convergence speed and accuracy. To bridge the gap, we introduce a novel layer-wise divergence analysis that uncovers the distinct update pattern of FO and ZO optimization. Aiming to resemble the learning capacity of FO method from the findings, we propose Divergence-driven Zeroth-Order (DiZO) optimization. DiZO conducts divergence-driven layer adaptation by incorporating projections to ZO updates, generating diverse-magnitude updates precisely scaled to layer-wise individual optimization needs. Our results demonstrate that DiZO significantly reduces the needed iterations for convergence without sacrificing throughput, cutting training GPU hours by up to 48% on various datasets. Moreover, DiZO consistently outperforms the representative ZO baselines in fine-tuning RoBERTa-large, OPT-series, and Llama-series on downstream tasks and, in some cases, even surpasses memory-intensive FO fine-tuning. Our code is released at https://anonymous.4open.science/r/DiZO-E86D.
Authors: Thomas Foster, Anya Sims, Johannes Forkel, Mattie Fellows, Jakob Foerster
Abstract: Reinforcement learning is now widely adopted as the final stage of large language model training, especially for reasoning-style tasks such as maths problems. Typically, models attempt each question many times during a single training step and attempt to learn from their successes and failures. However, we demonstrate that throughout training with two popular algorithms (PPO and VinePPO) on two widely used datasets, many questions are either solved by all attempts - meaning they are already learned - or by none - providing no meaningful training signal. To address this, we adapt a method from the reinforcement learning literature - sampling for learnability - and apply it to the reinforcement learning stage of LLM training. Our curriculum prioritises questions with high variance of success, i.e. those where the agent sometimes succeeds, but not always. Our findings demonstrate that this curriculum consistently boosts training performance across multiple algorithms and datasets, paving the way for more efficient and effective reinforcement learning with LLMs.
Authors: Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu, Yuhang Zhou, Te Gao, Wanxiang Che
Abstract: Recent advancements in reasoning with large language models (RLLMs), such as OpenAI-O1 and DeepSeek-R1, have demonstrated their impressive capabilities in complex domains like mathematics and coding. A central factor in their success lies in the application of long chain-of-thought (Long CoT) characteristics, which enhance reasoning abilities and enable the solution of intricate problems. However, despite these developments, a comprehensive survey on Long CoT is still lacking, limiting our understanding of its distinctions from traditional short chain-of-thought (Short CoT) and complicating ongoing debates on issues like "overthinking" and "inference-time scaling." This survey seeks to fill this gap by offering a unified perspective on Long CoT. (1) We first distinguish Long CoT from Short CoT and introduce a novel taxonomy to categorize current reasoning paradigms. (2) Next, we explore the key characteristics of Long CoT: deep reasoning, extensive exploration, and feasible reflection, which enable models to handle more complex tasks and produce more efficient, coherent outcomes compared to the shallower Short CoT. (3) We then investigate key phenomena such as the emergence of Long CoT with these characteristics, including overthinking, and inference-time scaling, offering insights into how these processes manifest in practice. (4) Finally, we identify significant research gaps and highlight promising future directions, including the integration of multi-modal reasoning, efficiency improvements, and enhanced knowledge frameworks. By providing a structured overview, this survey aims to inspire future research and further the development of logical reasoning in artificial intelligence.
Authors: Bingqian Lin, Yunshuang Nie, Khun Loun Zai, Ziming Wei, Mingfei Han, Rongtao Xu, Minzhe Niu, Jianhua Han, Liang Lin, Cewu Lu, Xiaodan Liang
Abstract: Building Vision-Language Navigation (VLN) agents which can navigate following natural language instructions is a long-standing goal in human-robot interaction applications. Recent studies have revealed the potential of training open-source Large Language Models (LLMs) to unleash LLMs' reasoning ability for improving navigation, and simultaneously mitigate the domain gap between LLMs' training corpus and the VLN task. However, these approaches primarily adopt direct input-output mapping paradigms, causing the mapping learning difficult and the navigational decisions unexplainable. Chain-of-Thought (CoT) training is a promising way to improve both navigational decision accuracy and interpretability, while the complexity of the navigation task makes the perfect CoT labels unavailable and may lead to overfitting through pure CoT supervised fine-tuning. In this paper, we propose a novel sElf-improving embodied reasoning framework for boosting LLM-based vision-language Navigation, dubbed EvolveNav. Our EvolveNav consists of two stages: (1) Formalized CoT Supervised Fine-Tuning, where we train the model with formalized CoT labels to both activate the model's navigational reasoning capabilities and increase the reasoning speed; (2) Self-Reflective Post-Training, where the model is iteratively trained with its own reasoning outputs as self-enriched CoT labels to enhance the supervision diversity. A self-reflective auxiliary task is also introduced to encourage learning correct reasoning patterns by contrasting with wrong ones. Experimental results on the popular VLN benchmarks demonstrate the superiority of EvolveNav over previous LLM-based VLN approaches. Code is available at https://github.com/expectorlin/EvolveNav.
Authors: Parshin Shojaee, Iman Mirzadeh, Keivan Alizadeh, Maxwell Horton, Samy Bengio, Mehrdad Farajtabar
Abstract: Recent generations of language models have introduced Large Reasoning Models (LRMs) that generate detailed thinking processes before providing answers. While these models demonstrate improved performance on reasoning benchmarks, their fundamental capabilities, scaling properties, and limitations remain insufficiently understood. Current evaluations primarily focus on established math and coding benchmarks, emphasizing final answer accuracy. However, this evaluation paradigm often suffers from contamination and does not provide insights into the reasoning traces. In this work, we systematically investigate these gaps with the help of controllable puzzle environments that allow precise manipulation of complexity while maintaining consistent logical structures. This setup enables the analysis of not only final answers but also the internal reasoning traces, offering insights into how LRMs think. Through extensive experiments, we show that LRMs face a complete accuracy collapse beyond certain complexities. Moreover, they exhibit a counterintuitive scaling limit: their reasoning effort increases with problem complexity up to a point, then declines despite having remaining token budget. By comparing LRMs with their standard LLM counterparts under same inference compute, we identify three performance regimes: (1) low-complexity tasks where standard models outperform LRMs, (2) medium-complexity tasks where LRMs demonstrates advantage, and (3) high-complexity tasks where both models face complete collapse. We found that LRMs have limitations in exact computation: they fail to use explicit algorithms and reason inconsistently across scales. We also investigate the reasoning traces in more depth, studying the patterns of explored solutions and analyzing the models' computational behavior, shedding light on their strengths, limitations, and raising questions about their reasoning capabilities.
Authors: Haoxuan Che, Haibo Jin, Zhengrui Guo, Yi Lin, Cheng Jin, Hao Chen
Abstract: LLMs have demonstrated significant potential in Medical Report Generation (MRG), yet their development requires large amounts of medical image-report pairs, which are commonly scattered across multiple centers. Centralizing these data is exceptionally challenging due to privacy regulations, thereby impeding model development and broader adoption of LLM-driven MRG models. To address this challenge, we present FedMRG, the first framework that leverages Federated Learning (FL) to enable privacy-preserving, multi-center development of LLM-driven MRG models, specifically designed to overcome the critical challenge of communication-efficient LLM training under multi-modal data heterogeneity. To start with, our framework tackles the fundamental challenge of communication overhead in FL-LLM tuning by employing low-rank factorization to efficiently decompose parameter updates, significantly reducing gradient transmission costs and making LLM-driven MRG feasible in bandwidth-constrained FL settings. Furthermore, we observed the dual heterogeneity in MRG under the FL scenario: varying image characteristics across medical centers, as well as diverse reporting styles and terminology preferences. To address this, we further enhance FedMRG with (1) client-aware contrastive learning in the MRG encoder, coupled with diagnosis-driven prompts, which capture both globally generalizable and locally distinctive features while maintaining diagnostic accuracy; and (2) a dual-adapter mutual boosting mechanism in the MRG decoder that harmonizes generic and specialized adapters to address variations in reporting styles and terminology. Through extensive evaluation of our established FL-MRG benchmark, we demonstrate the generalizability and adaptability of FedMRG, underscoring its potential in harnessing multi-center data and generating clinically accurate reports while maintaining communication efficiency.
Authors: Zhiting Mei, Christina Zhang, Tenny Yin, Justin Lidard, Ola Shorinwa, Anirudha Majumdar
Abstract: Reasoning language models have set state-of-the-art (SOTA) records on many challenging benchmarks, enabled by multi-step reasoning induced using reinforcement learning. However, like previous language models, reasoning models are prone to generating confident, plausible responses that are incorrect (hallucinations). Knowing when and how much to trust these models is critical to the safe deployment of reasoning models in real-world applications. To this end, we explore uncertainty quantification of reasoning models in this work. Specifically, we ask three fundamental questions: First, are reasoning models well-calibrated? Second, does deeper reasoning improve model calibration? Finally, inspired by humans' innate ability to double-check their thought processes to verify the validity of their answers and their confidence, we ask: can reasoning models improve their calibration by explicitly reasoning about their chain-of-thought traces? We introduce introspective uncertainty quantification (UQ) to explore this direction. In extensive evaluations on SOTA reasoning models across a broad range of benchmarks, we find that reasoning models: (i) are typically overconfident, with self-verbalized confidence estimates often greater than 85% particularly for incorrect responses, (ii) become even more overconfident with deeper reasoning, and (iii) can become better calibrated through introspection (e.g., o3-Mini and DeepSeek R1) but not uniformly (e.g., Claude 3.7 Sonnet becomes more poorly calibrated). Lastly, we conclude with important research directions to design necessary UQ benchmarks and improve the calibration of reasoning models.
Authors: Runcong Zhao, Artem Bobrov, Jiazheng Li, Yulan He
Abstract: Effective feedback is essential for student learning but is time-intensive for teachers. We present LearnLens, a modular, LLM-based system that generates personalised, curriculum-aligned feedback in science education. LearnLens comprises three components: (1) an error-aware assessment module that captures nuanced reasoning errors; (2) a curriculum-grounded generation module that uses a structured, topic-linked memory chain rather than traditional similarity-based retrieval, improving relevance and reducing noise; and (3) an educator-in-the-loop interface for customisation and oversight. LearnLens addresses key challenges in existing systems, offering scalable, high-quality feedback that empowers both teachers and students.
Authors: Ammar Ahmed, Ali Shariq Imran
Abstract: This systematic literature review examines the role of large language models (LLMs) in UI/UX design, synthesizing findings from 38 peer-reviewed studies published between 2022 and 2025. We identify key LLMs in use, including GPT-4, Gemini, and PaLM, and map their integration across the design lifecycle, from ideation to evaluation. Common practices include prompt engineering, human-in-the-loop workflows, and multimodal input. While LLMs are reshaping design processes, challenges such as hallucination, prompt instability, and limited explainability persist. Our findings highlight LLMs as emerging collaborators in design, and we propose directions for the ethical, inclusive, and effective integration of these technologies.
Authors: Eric Xing, Mingkai Deng, Jinyu Hou, Zhiting Hu
Abstract: World Model, the supposed algorithmic surrogate of the real-world environment which biological agents experience with and act upon, has been an emerging topic in recent years because of the rising needs to develop virtual agents with artificial (general) intelligence. There has been much debate on what a world model really is, how to build it, how to use it, and how to evaluate it. In this essay, starting from the imagination in the famed Sci-Fi classic Dune, and drawing inspiration from the concept of "hypothetical thinking" in psychology literature, we offer critiques of several schools of thoughts on world modeling, and argue the primary goal of a world model to be simulating all actionable possibilities of the real world for purposeful reasoning and acting. Building on the critiques, we propose a new architecture for a general-purpose world model, based on hierarchical, multi-level, and mixed continuous/discrete representations, and a generative and self-supervision learning framework, with an outlook of a Physical, Agentic, and Nested (PAN) AGI system enabled by such a model.
Authors: Ahmed Bahloul, Simon Malberg
Abstract: Modern language models address complex questions through chain-of-thought (CoT) reasoning (Wei et al., 2023) and retrieval augmentation (Lewis et al., 2021), yet struggle with error propagation and knowledge integration. Tree-structured reasoning methods, particularly the Probabilistic Tree-of-Thought (ProbTree)(Cao et al., 2023) framework, mitigate these issues by decomposing questions into hierarchical structures and selecting answers through confidence-weighted aggregation of parametric and retrieved knowledge (Yao et al., 2023). However, ProbTree's static implementation introduces two key limitations: (1) the reasoning tree is fixed during the initial construction phase, preventing dynamic adaptation to intermediate results, and (2) each node requires exhaustive evaluation of all possible solution strategies, creating computational inefficiency. We present a dynamic reinforcement learning (Sutton and Barto, 2018) framework that transforms tree-based reasoning into an adaptive process. Our approach incrementally constructs the reasoning tree based on real-time confidence estimates, while learning optimal policies for action selection (decomposition, retrieval, or aggregation). This maintains ProbTree's probabilistic rigor while improving both solution quality and computational efficiency through selective expansion and focused resource allocation. The work establishes a new paradigm for treestructured reasoning that balances the reliability of probabilistic frameworks with the flexibility required for real-world question answering systems.