Authors: Happymore Masoka
Abstract: Despite rapid advances in multilingual natural language processing (NLP), the Bantu language Shona remains under-served in terms of morphological analysis and language-aware tools. This paper presents Shona spaCy, an open-source, rule-based morphological pipeline for Shona built on the spaCy framework. The system combines a curated JSON lexicon with linguistically grounded rules to model noun-class prefixes (Mupanda 1-18), verbal subject concords, tense-aspect markers, ideophones, and clitics, integrating these into token-level annotations for lemma, part-of-speech, and morphological features. The toolkit is available via pip install shona-spacy, with source code at https://github.com/HappymoreMasoka/shona-spacy and a PyPI release at https://pypi.org/project/shona-spacy/0.1.4/. Evaluation on formal and informal Shona corpora yields 90% POS-tagging accuracy and 88% morphological-feature accuracy, while maintaining transparency in its linguistic decisions. By bridging descriptive grammar and computational implementation, Shona spaCy advances NLP accessibility and digital inclusion for Shona speakers and provides a template for morphological analysis tools for other under-resourced Bantu languages.
URLs: https://github.com/HappymoreMasoka/shona-spacy, https://pypi.org/project/shona-spacy/0.1.4/.
Authors: Dong Liu, Yanxuan Yu
Abstract: Retrieval-Augmented Generation (RAG) systems have become a dominant approach to augment large language models (LLMs) with external knowledge. However, existing vector database (VecDB) retrieval pipelines rely on flat or single-resolution indexing structures, which cannot adapt to the varying semantic granularity required by diverse user queries. This limitation leads to suboptimal trade-offs between retrieval speed and contextual relevance. To address this, we propose \textbf{Semantic Pyramid Indexing (SPI)}, a novel multi-resolution vector indexing framework that introduces query-adaptive resolution control for RAG in VecDBs. Unlike existing hierarchical methods that require offline tuning or separate model training, SPI constructs a semantic pyramid over document embeddings and dynamically selects the optimal resolution level per query through a lightweight classifier. This adaptive approach enables progressive retrieval from coarse-to-fine representations, significantly accelerating search while maintaining semantic coverage. We implement SPI as a plugin for both FAISS and Qdrant backends and evaluate it across multiple RAG tasks including MS MARCO, Natural Questions, and multimodal retrieval benchmarks. SPI achieves up to \textbf{5.7$\times$} retrieval speedup and \textbf{1.8$\times$} memory efficiency gain while improving end-to-end QA F1 scores by up to \textbf{2.5 points} compared to strong baselines. Our theoretical analysis provides guarantees on retrieval quality and latency bounds, while extensive ablation studies validate the contribution of each component. The framework's compatibility with existing VecDB infrastructures makes it readily deployable in production RAG systems. Code is availabe at \href{https://github.com/FastLM/SPI_VecDB}{https://github.com/FastLM/SPI\_VecDB}.
URLs: https://github.com/FastLM/SPI_VecDB, https://github.com/FastLM/SPI\_VecDB
Authors: Linus Stuhlmann, Mauricio Fadel Argerich, Jonathan F\"urst
Abstract: Running large language models (LLMs) locally is becoming increasingly common. While the growing availability of small open-source models and inference engines has lowered the entry barrier, users now face an overwhelming number of configuration choices. Identifying an optimal configuration -- balancing functional and non-functional requirements -- requires substantial manual effort. While several benchmarks target LLM inference, they are designed for narrow evaluation goals and not user-focused. They fail to integrate relevant system and task-specific metrics into a unified, easy-to-use benchmark that supports multiple inference engines, usage scenarios, and quantization levels. To address this gap, we present Bench360 -- Benchmarking Local LLM Inference from 360{\deg}. Bench360 allows users to easily define their own custom tasks along with datasets and relevant task-specific metrics and then automatically benchmarks selected LLMs, inference engines, and quantization levels across different usage scenarios (single stream, batch & server). Bench360 tracks a wide range of metrics, including (1) system metrics -- such as Computing Performance (e.g., latency, throughput), Resource Usage (e.g., energy per query), and Deployment (e.g., cold start time) -- and (2) task-specific metrics such as ROUGE, F1 score or accuracy. We demonstrate Bench360 on four common LLM tasks -- General Knowledge & Reasoning, QA, Summarization and Text-to-SQL -- across three hardware platforms and four state of the art inference engines. Our results reveal several interesting trade-offs between task performance and system-level efficiency, highlighting the differences in inference engines and models. Most importantly, there is no single best setup for local inference, which strongly motivates the need for a framework such as Bench360.
Authors: Mohamed Mahdi
Abstract: Large Language Models (LLMs) are the engines driving today's AI agents. The better these models understand human languages, the more natural and user-friendly the interaction with AI becomes, from everyday devices like computers and smartwatches to any tool that can act intelligently. Yet, the ability of industrial-scale LLMs to comprehend low-resource languages, such as Tunisian Arabic (Tunizi), is often overlooked. This neglect risks excluding millions of Tunisians from fully interacting with AI in their own language, pushing them toward French or English. Such a shift not only threatens the preservation of the Tunisian dialect but may also create challenges for literacy and influence younger generations to favor foreign languages. In this study, we introduce a novel dataset containing parallel Tunizi, standard Tunisian Arabic, and English translations, along with sentiment labels. We benchmark several popular LLMs on three tasks: transliteration, translation, and sentiment analysis. Our results reveal significant differences between models, highlighting both their strengths and limitations in understanding and processing Tunisian dialects. By quantifying these gaps, this work underscores the importance of including low-resource languages in the next generation of AI systems, ensuring technology remains accessible, inclusive, and culturally grounded.
Authors: Yuetian Zou, Hanlei Zhang, Hua Xu, Songze Li, Long Xiao
Abstract: Textual open intent classification is crucial for real-world dialogue systems, enabling robust detection of unknown user intents without prior knowledge and contributing to the robustness of the system. While adaptive decision boundary methods have shown great potential by eliminating manual threshold tuning, existing approaches assume isotropic distributions of known classes, restricting boundaries to balls and overlooking distributional variance along different directions. To address this limitation, we propose EliDecide, a novel method that learns ellipsoid decision boundaries with varying scales along different feature directions. First, we employ supervised contrastive learning to obtain a discriminative feature space for known samples. Second, we apply learnable matrices to parameterize ellipsoids as the boundaries of each known class, offering greater flexibility than spherical boundaries defined solely by centers and radii. Third, we optimize the boundaries via a novelly designed dual loss function that balances empirical and open-space risks: expanding boundaries to cover known samples while contracting them against synthesized pseudo-open samples. Our method achieves state-of-the-art performance on multiple text intent benchmarks and further on a question classification dataset. The flexibility of the ellipsoids demonstrates superior open intent detection capability and strong potential for generalization to more text classification tasks in diverse complex open-world scenarios.
Authors: Giulio Antonio Abbo, Tony Belpaeme
Abstract: Large language models are increasingly used in applications where alignment with human values is critical. While model fine-tuning is often employed to ensure safe responses, this technique is static and does not lend itself to everyday situations involving dynamic values and preferences. In this paper, we present a practical, reproducible, and model-agnostic procedure to evaluate whether a prompt candidate can effectively steer generated text toward specific human values, formalising a scoring method to quantify the presence and gain of target values in generated responses. We apply our method to a variant of the Wizard-Vicuna language model, using Schwartz's theory of basic human values and a structured evaluation through a dialogue dataset. With this setup, we compare a baseline prompt to one explicitly conditioned on values, and show that value steering is possible even without altering the model or dynamically optimising prompts.
Authors: Samarth Garg, Deeksha Varshney, Divya Singh
Abstract: The rise of social networks has not only facilitated communication but also allowed the spread of harmful content. Although significant advances have been made in detecting toxic language in textual data, the exploration of concept-based explanations in toxicity detection remains limited. In this study, we leverage various subtype attributes present in toxicity detection datasets, such as obscene, threat, insult, identity attack, and sexual explicit as concepts that serve as strong indicators to identify whether language is toxic. However, disproportionate attribution of concepts towards the target class often results in classification errors. Our work introduces an interpretability technique based on the Concept Gradient (CG) method which provides a more causal interpretation by measuring how changes in concepts directly affect the output of the model. This is an extension of traditional gradient-based methods in machine learning, which often focus solely on input features. We propose the curation of Targeted Lexicon Set, which captures toxic words that contribute to misclassifications in text classification models. To assess the significance of these lexicon sets in misclassification, we compute Word-Concept Alignment (WCA) scores, which quantify the extent to which these words lead to errors due to over-attribution to toxic concepts. Finally, we introduce a lexicon-free augmentation strategy by generating toxic samples that exclude predefined toxic lexicon sets. This approach allows us to examine whether over-attribution persists when explicit lexical overlap is removed, providing insights into the model's attribution on broader toxic language patterns.
Authors: Saleh Almohaimeed, Saad Almohaimeed, Mousa Jari, Khaled A. Alobaid, Fahad Alotaibi
Abstract: Many AI detection models have been developed to counter the presence of articles created by artificial intelligence (AI). However, if a human-authored article is slightly polished by AI, a shift will occur in the borderline decision of these AI detection models, leading them to consider it AI-generated article. This misclassification may result in falsely accusing authors of AI plagiarism and harm the credibility of AI detector models. In English, some efforts were made to meet this challenge, but not in Arabic. In this paper, we generated two datasets. The first dataset contains 800 Arabic articles, half AI-generated and half human-authored. We used it to evaluate 14 Large Language models (LLMs) and commercial AI detectors to assess their ability in distinguishing between human-authored and AI-generated articles. The best 8 models were chosen to act as detectors for our primary concern, which is whether they would consider slightly polished human text as AI-generated. The second dataset, Ar-APT, contains 400 Arabic human-authored articles polished by 10 LLMs using 4 polishing settings, totaling 16400 samples. We use it to evaluate the 8 nominated models and determine whether slight polishing will affect their performance. The results reveal that all AI detectors incorrectly attribute a significant number of articles to AI. The best performing LLM, Claude-4 Sonnet, achieved 83.51%, their performance decreased to 57.63% for articles slightly polished by LLaMA-3. Whereas for the best performing commercial model, originality.AI, that achieves 92% accuracy, dropped to 12% for articles slightly polished by Mistral or Gemma-3.
Authors: Boyang Zhou, Johan Lindqvist, Lindsey Li
Abstract: We reproduce the central claims of Test-Time Training on Nearest Neighbors for Large Language Models (Hardt and Sun, 2024), which proposes adapting a language model at inference time by fine-tuning on retrieved nearest-neighbor sequences. Using pretrained RoBERTa embeddings indexed with Faiss, we retrieve 20 neighbors per test input and apply one gradient update per neighbor across GPT-2 (117M, 774M), GPT-Neo (1.3B), and R1-Distilled-Qwen2.5-1.5B. Our experiments confirm that test-time training significantly reduces perplexity and bits-per-byte metrics across diverse domains from The Pile, with the largest improvements in structured or specialized datasets such as GitHub and EuroParl. We further validate that models not pretrained on The Pile benefit more from this adaptation than models already trained on similar data, allowing smaller models to approach the performance of larger ones. Due to infrastructure limitations, we introduce a memory-efficient retrieval implementation that loads only required line offsets rather than entire files, reducing RAM requirements from over 128 GB per server to 32 GB. We also extend the original study by evaluating R1-Distilled-Qwen2.5-1.5B, showing that test-time training yields consistent gains even for modern reasoning-optimized architectures. Overall, our results support the robustness and generality of nearest-neighbor test-time training while highlighting practical considerations for reproducing large-scale retrieval-augmented adaptation.
Authors: JaeSeong Kim, Suan Lee
Abstract: Multilingual LLMs demonstrate strong performance across diverse languages, yet there has been limited systematic analysis of how language information is structured within their internal representation space and how it emerges across layers. We conduct a comprehensive probing study on six multilingual LLMs, covering all 268 transformer layers, using linear and nonlinear probes together with a new Token--Language Alignment analysis to quantify the layer-wise dynamics and geometric structure of language encoding. Our results show that language information becomes sharply separated in the first transformer block (+76.4$\pm$8.2 percentage points from Layer 0 to 1) and remains almost fully linearly separable throughout model depth. We further find that the alignment between language directions and vocabulary embeddings is strongly tied to the language composition of the training data. Notably, Chinese-inclusive models achieve a ZH Match@Peak of 16.43\%, whereas English-centric models achieve only 3.90\%, revealing a 4.21$\times$ structural imprinting effect. These findings indicate that multilingual LLMs distinguish languages not by surface script features but by latent representational structures shaped by the training corpus. Our analysis provides practical insights for data composition strategies and fairness in multilingual representation learning. All code and analysis scripts are publicly available at: https://github.com/thisiskorea/How-Language-Directions-Align-with-Token-Geometry-in-Multilingual-LLMs.
Authors: Jonathon Dilworth, Hui Yang, Jiaoyan Chen, Yongsheng Gao
Abstract: SNOMED CT is a biomedical ontology with a hierarchical representation of large-scale concepts. Knowledge retrieval in SNOMED CT is critical for its application, but often proves challenging due to language ambiguity, synonyms, polysemies and so on. This problem is exacerbated when the queries are out-of-vocabulary (OOV), i.e., having no equivalent matchings in the ontology. In this work, we focus on the problem of hierarchical concept retrieval from SNOMED CT with OOV queries, and propose an approach based on language model-based ontology embeddings. For evaluation, we construct OOV queries annotated against SNOMED CT concepts, testing the retrieval of the most direct subsumers and their less relevant ancestors. We find that our method outperforms the baselines including SBERT and two lexical matching methods. While evaluated against SNOMED CT, the approach is generalisable and can be extended to other ontologies. We release code, tools, and evaluation datasets at https://github.com/jonathondilworth/HR-OOV.
Authors: Juan P. Cadile
Abstract: We investigate empathy-in-action -- the willingness to sacrifice task efficiency to address human needs -- as a linear direction in LLM activation space. Using contrastive prompts grounded in the Empathy-in-Action (EIA) benchmark, we test detection and steering across Phi-3-mini-4k (3.8B), Qwen2.5-7B (safety-trained), and Dolphin-Llama-3.1-8B (uncensored). Detection: All models show AUROC 0.996-1.00 at optimal layers. Uncensored Dolphin matches safety-trained models, demonstrating empathy encoding emerges independent of safety training. Phi-3 probes correlate strongly with EIA behavioral scores (r=0.71, p<0.01). Cross-model probe agreement is limited (Qwen: r=-0.06, Dolphin: r=0.18), revealing architecture-specific implementations despite convergent detection. Steering: Qwen achieves 65.3% success with bidirectional control and coherence at extreme interventions. Phi-3 shows 61.7% success with similar coherence. Dolphin exhibits asymmetric steerability: 94.4% success for pro-empathy steering but catastrophic breakdown for anti-empathy (empty outputs, code artifacts). Implications: The detection-steering gap varies by model. Qwen and Phi-3 maintain bidirectional coherence; Dolphin shows robustness only for empathy enhancement. Safety training may affect steering robustness rather than preventing manipulation, though validation across more models is needed.
Authors: Hossain Shaikh Saadi, Faria Alam, Mario Sanz-Guerrero, Minh Duc Bui, Manuel Mager, Katharina von der Wense
Abstract: This paper presents JGU Mainz's winning system for the BLP-2025 Shared Task on Code Generation from Bangla Instructions. We propose a multi-agent-based pipeline. First, a code-generation agent produces an initial solution from the input instruction. The candidate program is then executed against the provided unit tests (pytest-style, assert-based). Only the failing cases are forwarded to a debugger agent, which reruns the tests, extracts error traces, and, conditioning on the error messages, the current program, and the relevant test cases, generates a revised solution. Using this approach, our submission achieved first place in the shared task with a $Pass@1$ score of 95.4. We also make our code public.
Authors: Michael Carl, Takanori Mizowaki, Aishvarya Raj, Masaru Yamada, Devi Sri Bandaru, Yuxiang Wei, Xinyue Ren
Abstract: Building on the Extended Mind (EM) theory and radical enactivism, this article suggests an alternative to representation-based models of the mind. We lay out a novel ABC framework of the translating mind, in which translation is not the manipulation of static interlingual correspondences but an enacted activity, dynamically integrating affective, behavioral, and cognitive (ABC) processes. Drawing on Predictive Processing and (En)Active Inference, we argue that the translator's mind emerges, rather than being merely extended, through loops of brain-body-environment interactions. This non-representational account reframes translation as skillful participation in sociocultural practice, where meaning is co-created in real time through embodied interaction with texts, tools, and contexts.
Authors: Eva Neu, Brian Dillon, Katrin Erk
Abstract: Intransitive verbs fall into two different syntactic classes, unergatives and unaccusatives. It has long been argued that verbs describing an agentive action are more likely to appear in an unergative syntax, and those describing a telic event to appear in an unaccusative syntax. However, recent work by Kim et al. (2024) found that human ratings for agentivity and telicity were a poor predictor of the syntactic behavior of intransitives. Here we revisit this question using interpretable dimensions, computed from seed words on opposite poles of the agentive and telic scales. Our findings support the link between unergativity/unaccusativity and agentivity/telicity, and demonstrate that using interpretable dimensions in conjunction with human judgments can offer valuable evidence for semantic properties that are not easily evaluated in rating tasks.
Authors: Oscar Chew, Po-Yi Lu, Jayden Lin, Kuan-Hao Huang, Hsuan-Tien Lin
Abstract: Recent studies show that text to image (T2I) diffusion models are vulnerable to backdoor attacks, where a trigger in the input prompt can steer generation toward harmful or unintended content. To address this, we introduce PEPPER (PErcePtion Guided PERturbation), a backdoor defense that rewrites the caption into a semantically distant yet visually similar caption while adding unobstructive elements. With this rewriting strategy, PEPPER disrupt the trigger embedded in the input prompt, dilute the influence of trigger tokens and thereby achieve enhanced robustness. Experiments show that PEPPER is particularly effective against text encoder based attacks, substantially reducing attack success while preserving generation quality. Beyond this, PEPPER can be paired with any existing defenses yielding consistently stronger and generalizable robustness than any standalone method. Our code will be released on Github.
Authors: Seyed Mohssen Ghafari, Ronny Kol, Juan C. Quiroz, Nella Luan, Monika Patial, Chanaka Rupasinghe, Herman Wandabwa, Luiz Pizzato
Abstract: Large language models (LLMs) frequently generate responses that are lengthy and verbose, filled with redundant or unnecessary details. This diminishes clarity and user satisfaction, and it increases costs for model developers, especially with well-known proprietary models that charge based on the number of output tokens. In this paper, we introduce a novel reference-free metric for evaluating the conciseness of responses generated by LLMs. Our method quantifies non-essential content without relying on gold standard references and calculates the average of three calculations: i) a compression ratio between the original response and an LLM abstractive summary; ii) a compression ratio between the original response and an LLM extractive summary; and iii) wordremoval compression, where an LLM removes as many non-essential words as possible from the response while preserving its meaning, with the number of tokens removed indicating the conciseness score. Experimental results demonstrate that our proposed metric identifies redundancy in LLM outputs, offering a practical tool for automated evaluation of response brevity in conversational AI systems without the need for ground truth human annotations.
Authors: Kang Wang, Xiangyu Duan, Tianyi Du
Abstract: Unlike human reasoning in abstract conceptual spaces, large language models (LLMs) typically reason by generating discrete tokens, which potentially limit their expressive power. The recent work Soft Thinking has shown that LLMs' latent reasoning via soft concepts is a promising direction, but LLMs are trained on discrete tokens. To reduce this gap between the soft concepts in reasoning and the discrete tokens in training, we propose Soft Concept Mixing (SCM), a soft concept aware training scheme that directly exposes the model to soft representations during training. Specifically, SCM constructs a soft concept vector by forming a probability-weighted average of embeddings. Then, this vector is mixed into the model's hidden states, which embody rich contextual information. Finally, the entire latent reasoning process is optimized with Reinforcement Learning (RL). Experiments on five reasoning benchmarks demonstrate that SCM improves the reasoning performance of LLMs, and simultaneously maintains a stable training dynamic.
Authors: Arip Asadulaev, Rayan Banerjee, Fakhri Karray, Martin Takac
Abstract: Recently, it was shown that small, looped architectures, such as Tiny Recursive Models (TRMs), can outperform Large Language Models (LLMs) on complex reasoning tasks, including the Abstraction and Reasoning Corpus (ARC). In this work, we investigate a core question: how can we further improve the efficiency of these methods with minimal changes? To address this, we frame the latent reasoning of TRMs as a form of classifier-free guidance and implicit policy improvement algorithm. Building on these insights, we propose a novel training scheme that provides a target for each loop during training. We demonstrate that our approach significantly enhances training efficiency. Our method reduces the total number of forward passes by 18x and eliminates halting mechanisms, while maintaining quality comparable to standard TRMs. Notably, we achieve 24% accuracy on ARC-1 with only 0.8M parameters, outperforming most LLMs.
Authors: Tatsuya Aoyama, Ethan Gotlieb Wilcox, Nathan Schneider
Abstract: Arguably, specialized attention heads dubbed induction heads (IHs) underlie the remarkable in-context learning (ICL) capabilities of modern language models (LMs); yet, a precise characterization of their formation remains unclear. In this study, we investigate the relationship between statistical properties of training data (for both natural and synthetic data) and IH formation. We show that (1) a simple equation combining batch size and context size predicts the point at which IHs form; (2) surface bigram repetition frequency and reliability strongly affect the formation of IHs, and we find a precise Pareto frontier in terms of these two values; and (3) local dependency with high bigram repetition frequency and reliability is sufficient for IH formation, but when the frequency and reliability are low, categoriality and the shape of the marginal distribution matter.
Authors: An Quang Tang, Xiuzhen Zhang, Minh Ngoc Dinh, Zhuang Li
Abstract: Online conversations have become more prevalent on public discussion platforms (e.g. Reddit). With growing controversial topics, it is desirable to summarize not only diverse arguments, but also their rationale and justification. Early studies on text summarization focus on capturing general salient information in source documents, overlooking the argumentative nature of online conversations. Recent research on conversation summarization although considers the argumentative relationship among sentences, fail to explicate deeper argument structure within sentences for summarization. In this paper, we propose a novel task of argument-aware quantitative summarization to reveal the claim-reason structure of arguments in conversations, with quantities measuring argument strength. We further propose ARQUSUMM, a novel framework to address the task. To reveal the underlying argument structure within sentences, ARQUSUMM leverages LLM few-shot learning grounded in the argumentation theory to identify propositions within sentences and their claim-reason relationships. For quantitative summarization, ARQUSUMM employs argument structure-aware clustering algorithms to aggregate arguments and quantify their support. Experiments show that ARQUSUMM outperforms existing conversation and quantitative summarization models and generate summaries representing argument structures that are more helpful to users, of high textual quality and quantification accuracy.
Authors: Junjie Hao, Chun Wang, Ying Qiao, Qiuyue Zuo, Qiya Song, Hua Ma, Xieping Gao
Abstract: Large language models and knowledge graphs offer strong potential for advancing research on historical culture by supporting the extraction, analysis, and interpretation of cultural heritage. Using Hunan's modern historical celebrities shaped by Huxiang culture as a case study, pre-trained large models can help researchers efficiently extract key information, including biographical attributes, life events, and social relationships, from textual sources and construct structured knowledge graphs. However, systematic data resources for Hunan's historical celebrities remain limited, and general-purpose models often underperform in domain knowledge extraction and structured output generation in such low-resource settings. To address these issues, this study proposes a supervised fine-tuning approach for enhancing domain-specific information extraction. First, we design a fine-grained, schema-guided instruction template tailored to the Hunan historical celebrities domain and build an instruction-tuning dataset to mitigate the lack of domain-specific training corpora. Second, we apply parameter-efficient instruction fine-tuning to four publicly available large language models - Qwen2.5-7B, Qwen3-8B, DeepSeek-R1-Distill-Qwen-7B, and Llama-3.1-8B-Instruct - and develop evaluation criteria for assessing their extraction performance. Experimental results show that all models exhibit substantial performance gains after fine-tuning. Among them, Qwen3-8B achieves the strongest results, reaching a score of 89.3866 with 100 samples and 50 training iterations. This study provides new insights into fine-tuning vertical large language models for regional historical and cultural domains and highlights their potential for cost-effective applications in cultural heritage knowledge extraction and knowledge graph construction.
Authors: Gyuwon Park
Abstract: Recent advances in vision-language models (VLMs) have enabled impressive multi-modal reasoning and understanding. Yet, whether these models truly grasp visual persuasion-how visual cues shape human attitudes and decisions-remains unclear. To probe this question, we construct a high-consensus dataset for binary persuasiveness judgment and introduce the taxonomy of Visual Persuasive Factors (VPFs), encompassing low-level perceptual, mid-level compositional, and high-level semantic cues. We also explore cognitive steering and knowledge injection strategies for persuasion-relevant reasoning. Empirical analysis across VLMs reveals a recall-oriented bias-models over-predict high persuasiveness-and weak discriminative power for low/mid-level features. In contrast, high-level semantic alignment between message and object presence emerges as the strongest predictor of human judgment. Among intervention strategies, simple instruction or unguided reasoning scaffolds yield marginal or negative effects, whereas concise, object-grounded rationales significantly improve precision and F1 scores. These results indicate that VLMs core limitation lies not in recognizing persuasive objects but in linking them to communicative intent.
Authors: Yunsung Kim, Mike Hardy, Joseph Tey, Candace Thille, Chris Piech
Abstract: AI-driven automated scoring systems offer scalable and efficient means of evaluating complex student-generated responses. Yet, despite increasing demand for transparency and interpretability, the field has yet to develop a widely accepted solution for interpretable automated scoring to be used in large-scale real-world assessments. This work takes a principled approach to address this challenge. We analyze the needs and potential benefits of interpretable automated scoring for various assessment stakeholders and develop four principles of interpretability -- Faithfulness, Groundedness, Traceability, and Interchangeability (FGTI) -- targeted at those needs. To illustrate the feasibility of implementing these principles, we develop the AnalyticScore framework for short answer scoring as a baseline reference framework for future research. AnalyticScore operates by (1) extracting explicitly identifiable elements of the responses, (2) featurizing each response into human-interpretable values using LLMs, and (3) applying an intuitive ordinal logistic regression model for scoring. In terms of scoring accuracy, AnalyticScore outperforms many uninterpretable scoring methods, and is within only 0.06 QWK of the uninterpretable SOTA on average across 10 items from the ASAP-SAS dataset. By comparing against human annotators conducting the same featurization task, we further demonstrate that the featurization behavior of AnalyticScore aligns well with that of humans.
Authors: J\'er\'emie Dentan, Alexi Canesse, Davide Buscaldi, Aymen Shabou, Sonia Vanier
Abstract: Claim-level Uncertainty Quantification (UQ) is a promising approach to mitigate the lack of reliability in Large Language Models (LLMs). We introduce MUCH, the first claim-level UQ benchmark designed for fair and reproducible evaluation of future methods under realistic conditions. It includes 4,873 samples across four European languages (English, French, Spanish, and German) and four instruction-tuned open-weight LLMs. Unlike prior claim-level benchmarks, we release 24 generation logits per token, facilitating the development of future white-box methods without re-generating data. Moreover, in contrast to previous benchmarks that rely on manual or LLM-based segmentation, we propose a new deterministic algorithm capable of segmenting claims using as little as 0.2% of the LLM generation time. This makes our segmentation approach suitable for real-time monitoring of LLM outputs, ensuring that MUCH evaluates UQ methods under realistic deployment constraints. Finally, our evaluations show that current methods still have substantial room for improvement in both performance and efficiency.
Authors: Quentin Anthony, Yury Tokpanov, Skyler Szot, Srivatsan Rajagopal, Praneeth Medepalli, Rishi Iyer, Vasu Shyam, Anna Golubeva, Ansh Chaurasia, Xiao Yang, Tomas Figliolia, Robert Washbourne, Drew Thorstensen, Amartey Pearson, Zack Grossbart, Jason van Patten, Emad Barsoum, Zhenyu Gu, Yao Fu, Beren Millidge
Abstract: We report on the first large-scale mixture-of-experts (MoE) pretraining study on pure AMD hardware, utilizing both MI300X GPUs with Pollara interconnect. We distill practical guidance for both systems and model design. On the systems side, we deliver a comprehensive cluster and networking characterization: microbenchmarks for all core collectives (all-reduce, reduce-scatter, all-gather, broadcast) across message sizes and GPU counts on Pollara. To our knowledge, this is the first at this scale. We further provide MI300X microbenchmarks on kernel sizing and memory bandwidth to inform model design. On the modeling side, we introduce and apply MI300X-aware transformer sizing rules for attention and MLP blocks and justify MoE widths that jointly optimize training throughput and inference latency. We describe our training stack in depth, including often-ignored utilities such as fault-tolerance and checkpoint-reshaping, as well as detailed information on our training recipe. We also provide a preview of our model architecture and base model - ZAYA1 (760M active, 8.3B total parameters MoE) - which will be further improved upon in forthcoming papers. ZAYA1-base achieves performance comparable to leading base models such as Qwen3-4B and Gemma3-12B at its scale and larger, and outperforms models including Llama-3-8B and OLMoE across reasoning, mathematics, and coding benchmarks. Together, these results demonstrate that the AMD hardware, network, and software stack are mature and optimized enough for competitive large-scale pretraining.
Authors: Yeqin Zhang, Yizheng Zhao, Chen Hu, Binxing Jiao, Daxin Jiang, Ruihang Miao, Cam-Tu Nguyen
Abstract: Text representation plays a critical role in tasks like clustering, retrieval, and other downstream applications. With the emergence of large language models (LLMs), there is increasing interest in harnessing their capabilities for this purpose. However, most of the LLMs are inherently causal and optimized for next-token prediction, making them suboptimal for producing holistic representations. To address this, recent studies introduced pretext tasks to adapt LLMs for text representation. Most of these tasks, however, rely on token-level prediction objectives, such as the masked next-token prediction (MNTP) used in LLM2Vec. In this work, we explore the untapped potential of context compression as a pretext task for unsupervised adaptation of LLMs. During compression pre-training, the model learns to generate compact memory tokens, which substitute the whole context for downstream sequence prediction. Experiments demonstrate that a well-designed compression objective can significantly enhance LLM-based text representations, outperforming models trained with token-level pretext tasks. Further improvements through contrastive learning produce a strong representation model (LLM2Comp) that outperforms contemporary LLM-based text encoders on a wide range of tasks while being more sample-efficient, requiring significantly less training data.
Authors: Diego Velazquez, Mikaela Grace, Konstantinos Karageorgos, Lawrence Carin, Aaron Schliem, Dimitrios Zaikis, Roger Wechsler
Abstract: Automatic post-editing (APE) aims to correct errors in machine-translated text, enhancing translation quality, while reducing the need for human intervention. Despite advances in neural machine translation (NMT), the development of effective APE systems has been hindered by the lack of large-scale multilingual datasets specifically tailored to NMT outputs. To address this gap, we present and release LangMark, a new human-annotated multilingual APE dataset for English translation to seven languages: Brazilian Portuguese, French, German, Italian, Japanese, Russian, and Spanish. The dataset has 206,983 triplets, with each triplet consisting of a source segment, its NMT output, and a human post-edited translation. Annotated by expert human linguists, our dataset offers both linguistic diversity and scale. Leveraging this dataset, we empirically show that Large Language Models (LLMs) with few-shot prompting can effectively perform APE, improving upon leading commercial and even proprietary machine translation systems. We believe that this new resource will facilitate the future development and evaluation of APE systems.
Authors: Piotr P\k{e}zik, Filip \.Zarnecki, Konrad Kaczy\'nski, Anna Cichosz, Zuzanna Deckert, Monika Garnys, Izabela Grabarczyk, Wojciech Janowski, Sylwia Karasi\'nska, Aleksandra Kujawiak, Piotr Misztela, Maria Szyma\'nska, Karolina Walkusz, Igor Siek, Maciej Chrab\k{a}szcz, Anna Ko{\l}os, Agnieszka Karli\'nska, Karolina Seweryn, Aleksandra Krasnod\k{e}bska, Paula Betscher, Zofia Cie\'sli\'nska, Katarzyna Kowol, Artur Wilczek, Maciej Trzci\'nski, Katarzyna Dziewulska, Roman Roszko, Tomasz Berna\'s, Jurgita Vai\v{c}enonien\.e, Danuta Roszko, Pawe{\l} Levchuk, Pawe{\l} Kowalski, Irena Prawdzic-Jankowska, Marek Koz{\l}owski, S{\l}awomir Dadas, Rafa{\l} Po\'swiata, Alina Wr\'oblewska, Katarzyna Krasnowska-Kiera\'s, Maciej Ogrodniczuk, Micha{\l} Rudolf, Piotr Rybak, Karolina Saputa, Joanna Wo{\l}oszyn, Marcin Oleksy, Bart{\l}omiej Koptyra, Teddy Ferdinan, Stanis{\l}aw Wo\'zniak, Maciej Piasecki, Pawe{\l} Walkowiak, Konrad Wojtasik, Arkadiusz Janz, Przemys{\l}aw Kazienko, Julia Moska, Jan Koco\'n
Abstract: This paper describes the instruction dataset used to fine-tune a set of transformer-based large language models (LLMs) developed in the PLLuM (Polish Large Language Model) project. We present a functional typology of the organic, converted, and synthetic instructions used in PLLuM and share some observations about the implications of using human-authored versus synthetic instruction datasets in the linguistic adaptation of base LLMs. Additionally, we release the first representative subset of the PLLuM instruction corpus (PLLuMIC), which we believe to be useful in guiding and planning the development of similar datasets for other LLMs.
Authors: Vy Nguyen, Ziqi Xu, Jeffrey Chan, Estrid He, Feng Xia, Xiuzhen Zhang
Abstract: Large Language Models (LLMs) often produce fluent but factually incorrect responses, a phenomenon known as hallucination. Abstention, where the model chooses not to answer and instead outputs phrases such as "I don't know", is a common safeguard. However, existing abstention methods typically rely on post-generation signals, such as generation variations or feedback, which limits their ability to prevent unreliable responses in advance. In this paper, we introduce Aspect-Based Causal Abstention (ABCA), a new framework that enables early abstention by analysing the internal diversity of LLM knowledge through causal inference. This diversity reflects the multifaceted nature of parametric knowledge acquired from various sources, representing diverse aspects such as disciplines, legal contexts, or temporal frames. ABCA estimates causal effects conditioned on these aspects to assess the reliability of knowledge relevant to a given query. Based on these estimates, we enable two types of abstention: Type-1, where aspect effects are inconsistent (knowledge conflict), and Type-2, where aspect effects consistently support abstention (knowledge insufficiency). Experiments on standard benchmarks demonstrate that ABCA improves abstention reliability, achieves state-of-the-art performance, and enhances the interpretability of abstention decisions.
Authors: Mohammad Zare
Abstract: News text classification is a crucial task in natural language processing, essential for organizing and filtering the massive volume of digital content. Traditional methods typically rely on statistical features like term frequencies or TF-IDF values, which are effective at capturing word-level importance but often fail to reflect contextual meaning. In contrast, modern deep learning approaches utilize semantic features to understand word usage within context, yet they may overlook simple, high-impact statistical indicators. This paper introduces an Attention-Guided Feature Fusion (AGFF) model that combines statistical and semantic features in a unified framework. The model applies an attention-based mechanism to dynamically determine the relative importance of each feature type, enabling more informed classification decisions. Through evaluation on benchmark news datasets, the AGFF model demonstrates superior performance compared to both traditional statistical models and purely semantic deep learning models. The results confirm that strategic integration of diverse feature types can significantly enhance classification accuracy. Additionally, ablation studies validate the contribution of each component in the fusion process. The findings highlight the model's ability to balance and exploit the complementary strengths of statistical and semantic representations, making it a practical and effective solution for real-world news classification tasks.
Authors: Ziyang Wang, Yuanlei Zheng, Zhenbiao Cao, Xiaojin Zhang, Zhongyu Wei, Pei Fu, Zhenbo Luo, Wei Chen, Xiang Bai
Abstract: For industrial-scale text-to-SQL, supplying the entire database schema to Large Language Models (LLMs) is impractical due to context window limits and irrelevant noise. Schema linking, which filters the schema to a relevant subset, is therefore critical. However, existing methods incur prohibitive costs, struggle to trade off recall and noise, and scale poorly to large databases. We present \textbf{AutoLink}, an autonomous agent framework that reformulates schema linking as an iterative, agent-driven process. Guided by an LLM, AutoLink dynamically explores and expands the linked schema subset, progressively identifying necessary schema components without inputting the full database schema. Our experiments demonstrate AutoLink's superior performance, achieving state-of-the-art strict schema linking recall of \textbf{97.4\%} on Bird-Dev and \textbf{91.2\%} on Spider-2.0-Lite, with competitive execution accuracy, i.e., \textbf{68.7\%} EX on Bird-Dev (better than CHESS) and \textbf{34.9\%} EX on Spider-2.0-Lite (ranking 2nd on the official leaderboard). Crucially, AutoLink exhibits \textbf{exceptional scalability}, \textbf{maintaining high recall}, \textbf{efficient token consumption}, and \textbf{robust execution accuracy} on large schemas (e.g., over 3,000 columns) where existing methods severely degrade-making it a highly scalable, high-recall schema-linking solution for industrial text-to-SQL systems.
Authors: Tao Yuan, Haoli Bai, Yinfei Pan, Xuyang Cao, Tianyu Zhang, Lu Hou, Ting Hu, Xianzhi Yu
Abstract: With the increasing size of large language models, layer pruning has gained increased attention as a hardware-friendly approach for model compression. However, existing layer pruning methods struggle to simultaneously address key practical deployment challenges, including performance degradation, high training costs, and limited acceleration. To overcome these limitations, we propose \name, a task-\underline{E}ffective, training-\underline{E}conomical and inference-\underline{E}fficient layer pruning framework. \namespace introduces two key innovations: (1) a differentiable mask optimization method using a Gumbel-TopK sampler, enabling efficient and precise pruning mask search; and (2) an entropy-aware adaptive knowledge distillation strategy that enhances task performance. Extensive experiments over diverse model architectures and benchmarks demonstrate the superiority of our method over state-of-the-art approaches. Notably, \namespace achieves 96\% accuracy, a mere 0.8\% drop from the original model (96.8\%) on MATH-500 when pruning 25\% layers of Qwen3-32B, outperforming existing SOTA (95\%), with a 1.33$\times$ inference speedup by consuming merely 0.5B tokens (0.5\% of the post-training data volume).
Authors: Sizhe Zhou
Abstract: LLM-based conversational agents still struggle to maintain coherent, personalized interaction over many sessions: fixed context windows limit how much history can be kept in view, and most external memory approaches trade off between coarse retrieval over large chunks and fine-grained but fragmented views of the dialogue. Motivated by neo-Davidsonian event semantics, we propose an event-centric alternative that represents conversational history as short, event-like propositions which bundle together participants, temporal cues, and minimal local context, rather than as independent relation triples or opaque summaries. In contrast to work that aggressively compresses or forgets past content, our design aims to preserve information in a non-compressive form and make it more accessible, rather than more lossy. Concretely, we instruct an LLM to decompose each session into enriched elementary discourse units (EDUs) -- self-contained statements with normalized entities and source turn attributions -- and organize sessions, EDUs, and their arguments in a heterogeneous graph that supports associative recall. On top of this representation we build two simple retrieval-based variants that use dense similarity search and LLM filtering, with an optional graph-based propagation step to connect and aggregate evidence across related EDUs. Experiments on the LoCoMo and LongMemEval$_S$ benchmarks show that these event-centric memories match or surpass strong baselines, while operating with much shorter QA contexts. Our results suggest that structurally simple, event-level memory provides a principled and practical foundation for long-horizon conversational agents. Our code and data will be released at https://github.com/KevinSRR/EMem.
Authors: Yusuf \c{C}elebi, Mahmoud El Hussieni, \"Ozay Ezerceli
Abstract: This study presents PARROT (Persuasion and Agreement Robustness Rating of Output Truth), a robustness focused framework designed to measure the degradation in accuracy that occurs under social pressure exerted on users through authority and persuasion in large language models (LLMs) the phenomenon of sycophancy (excessive conformity). PARROT (i) isolates causal effects by comparing the neutral version of the same question with an authoritatively false version using a double-blind evaluation, (ii) quantifies confidence shifts toward the correct and imposed false responses using log-likelihood-based calibration tracking, and (iii) systematically classifies failure modes (e.g., robust correct, sycophantic agreement, reinforced error, stubborn error, self-correction, etc.) using an eight-state behavioral taxonomy. We evaluated 22 models using 1,302 MMLU-style multiple-choice questions across 13 domains and domain-specific authority templates. Findings show marked heterogeneity: advanced models (e.g., GPT-5, GPT-4.1, Claude Sonnet 4.5) exhibit low "follow rates" ($\leq 11\%$, GPT-5: 4\%) and minimal accuracy loss, while older/smaller models show severe epistemic collapse (GPT-4: 80\%, Qwen 2.5-1.5B: 94\%). The danger is not limited to response changes; weak models reduce confidence in the correct response while increasing confidence in the imposed incorrect response. While international law and global knowledge at the domain level exhibit high fragility, elementary mathematics is relatively resilient. Consequently, we argue that the goal of "resistance to overfitting pressure" should be addressed as a primary objective alongside accuracy, harm avoidance, and privacy for safe deployment in the real world.
Authors: Anshul Singh, Rohan Chaudhary, Gagneet Singh, Abhay Kumary
Abstract: The impressive performance of VLMs is largely measured on benchmarks that fail to capture the complexities of real-world scenarios. Existing datasets for tabular QA, such as WikiTableQuestions and FinQA, are overwhelmingly monolingual (English) and present tables in a digitally perfect, clean format. This creates a significant gap between research and practice. To address this, we present \textbf{MirageTVQA}, a new benchmark designed to evaluate VLMs on these exact dimensions. Featuring nearly 60,000 QA pairs across 24 languages, MirageTVQA challenges models with tables that are not only multilingual but also visually imperfect, incorporating realistic noise to mimic scanned documents. Our evaluation of the leading VLMs reveals two primary failure points: a severe degradation in performance (over 35\% drop for the best models) when faced with visual noise and a consistent English-first bias where reasoning abilities fail to transfer to other languages. MirageTVQA provides a benchmark for measuring and driving progress towards more robust VLM models for table reasoning. The dataset and the code are available at: https://github.com/anshulsc/MirageTVQA.
Authors: Benjamin White, Anastasia Shimorina
Abstract: Understanding and predicting user behavior on social media platforms is crucial for content recommendation and platform design. While existing approaches focus primarily on common actions like retweeting and liking, the prediction of rare but significant behaviors remains largely unexplored. This paper presents a hybrid methodology for social media user behavior prediction that addresses both frequent and infrequent actions across a diverse action vocabulary. We evaluate our approach on a large-scale Bluesky dataset containing 6.4 million conversation threads spanning 12 distinct user actions across 25 persona clusters. Our methodology combines four complementary approaches: (i) a lookup database system based on historical response patterns; (ii) persona-specific LightGBM models with engineered temporal and semantic features for common actions; (iii) a specialized hybrid neural architecture fusing textual and temporal representations for rare action classification; and (iv) generation of text replies. Our persona-specific models achieve an average macro F1-score of 0.64 for common action prediction, while our rare action classifier achieves 0.56 macro F1-score across 10 rare actions. These results demonstrate that effective social media behavior prediction requires tailored modeling strategies recognizing fundamental differences between action types. Our approach achieved first place in the SocialSim: Social-Media Based Personas challenge organized at the Social Simulation with LLMs workshop at COLM 2025.
Authors: Marii Ojastu, Hele-Andra Kuulmets, Aleksei Dorkin, Marika Borovikova, Dage S\"arg, Kairit Sirts
Abstract: In this paper, we present a localized and culturally adapted Estonian translation of the test set from the widely used commonsense reasoning benchmark, WinoGrande. We detail the translation and adaptation process carried out by translation specialists and evaluate the performance of both proprietary and open source models on the human translated benchmark. Additionally, we explore the feasibility of achieving high-quality machine translation by incorporating insights from the manual translation process into the design of a detailed prompt. This prompt is specifically tailored to address both the linguistic characteristics of Estonian and the unique translation challenges posed by the WinoGrande dataset. Our findings show that model performance on the human translated Estonian dataset is slightly lower than on the original English test set, while performance on machine-translated data is notably worse. Additionally, our experiments indicate that prompt engineering offers limited improvement in translation quality or model accuracy, and highlight the importance of involving language specialists in dataset translation and adaptation to ensure reliable and interpretable evaluations of language competency and reasoning in large language models.
Authors: Koena Ronny Mabokela, Tim Schlippe, Matthias W\"olfel
Abstract: Sentiment analysis can aid in understanding people's opinions and emotions on social issues. In multilingual communities sentiment analysis systems can be used to quickly identify social challenges in social media posts, enabling government departments to detect and address these issues more precisely and effectively. Recently, large-language models (LLMs) have become available to the wide public and initial analyses have shown that they exhibit magnificent zero-shot sentiment analysis abilities in English. However, there is no work that has investigated to leverage LLMs for sentiment analysis on social media posts in South African languages and detect social challenges. Consequently, in this work, we analyse the zero-shot performance of the state-of-the-art LLMs GPT-3.5, GPT-4, LlaMa 2, PaLM 2, and Dolly 2 to investigate the sentiment polarities of the 10 most emerging topics in English, Sepedi and Setswana social media posts that fall within the jurisdictional areas of 10 South African government departments. Our results demonstrate that there are big differences between the various LLMs, topics, and languages. In addition, we show that a fusion of the outcomes of different LLMs provides large gains in sentiment classification performance with sentiment classification errors below 1%. Consequently, it is now feasible to provide systems that generate reliable information about sentiment analysis to detect social challenges and draw conclusions about possible needs for actions on specific topics and within different language groups.
Authors: Mateusz Jacniacki, Mart\'i Carmona Serrat
Abstract: Conversational agents built on large language models (LLMs) are becoming increasingly prevalent, yet most systems are designed for one-on-one, turn-based exchanges rather than natural, asynchronous group chats. As AI assistants become widespread throughout digital platforms, from virtual assistants to customer service, developing natural and humanlike interaction patterns seems crucial for maintaining user trust and engagement. We present the Humanlike Multi-user Agent (HUMA), an LLM-based facilitator that participates in multi-party conversations using human-like strategies and timing. HUMA extends prior multi-user chatbot work with an event-driven architecture that handles messages, replies, reactions and introduces realistic response-time simulation. HUMA comprises three components-Router, Action Agent, and Reflection-which together adapt LLMs to group conversation dynamics. We evaluate HUMA in a controlled study with 97 participants in four-person role-play chats, comparing AI and human community managers (CMs). Participants classified CMs as human at near-chance rates in both conditions, indicating they could not reliably distinguish HUMA agents from humans. Subjective experience was comparable across conditions: community-manager effectiveness, social presence, and engagement/satisfaction differed only modestly with small effect sizes. Our results suggest that, in natural group chat settings, an AI facilitator can match human quality while remaining difficult to identify as nonhuman.
Authors: Xiaoyun Jin, Mirjam Ernestus, R. Harald Baayen
Abstract: We present a corpus-based investigation of how the pitch contours of monosyllabic words are realized in spontaneous conversational Mandarin, focusing on the effects of words' meanings. We used the generalized additive model to decompose a given observed pitch contour into a set of component pitch contours that are tied to different control variables and semantic predictors. Even when variables such as word duration, gender, speaker identity, tonal context, vowel height, and utterance position are controlled for, the effect of word remains a strong predictor of tonal realization. We present evidence that this effect of word is a semantic effect: word sense is shown to be a better predictor than word, and heterographic homophones are shown to have different pitch contours. The strongest evidence for the importance of semantics is that the pitch contours of individual word tokens can be predicted from their contextualized embeddings with an accuracy that substantially exceeds a permutation baseline. For phonetics, distributional semantics is a new kid on the block. Although our findings challenge standard theories of Mandarin tone, they fit well within the theoretical framework of the Discriminative Lexicon Model.
Authors: Daniil Ignatev, Ayman Santeer, Albert Gatt, Denis Paperno
Abstract: We propose a zero-shot method for Natural Language Inference (NLI) that leverages multimodal representations by grounding language in visual contexts. Our approach generates visual representations of premises using text-to-image models and performs inference by comparing these representations with textual hypotheses. We evaluate two inference techniques: cosine similarity and visual question answering. Our method achieves high accuracy without task-specific fine-tuning, demonstrating robustness against textual biases and surface heuristics. Additionally, we design a controlled adversarial dataset to validate the robustness of our approach. Our findings suggest that leveraging visual modality as a meaning representation provides a promising direction for robust natural language understanding.
Authors: Sajad Movahedi, Timur Carstensen, Arshia Afzal, Frank Hutter, Antonio Orvieto, Volkan Cevher
Abstract: Position information is essential for language modeling. In softmax transformers, Rotary Position Embeddings (\textit{RoPE}) encode positions through \textit{fixed-angle} rotations, while in linear transformers, order is handled via input-dependent (selective) gating that decays past key-value associations. Selectivity has generally been shown to improve language-related tasks. Inspired by this, we introduce \textit{Selective RoPE}, an \textit{input-dependent} rotary embedding mechanism, that generalizes \textit{RoPE}, and enables rotation in \textit{arbitrary angles} for both linear and softmax transformers. We show that softmax attention already performs a hidden form of these rotations on query-key pairs, uncovering an implicit positional structure. We further show that in state-space models and gated linear transformers, the real part manages forgetting while the imaginary part encodes positions through rotations. We validate our method by equipping gated transformers with \textit{Selective RoPE}, demonstrating that its input-dependent rotations improve performance in language modeling and on difficult sequence tasks like copying, state tracking, and retrieval.
Authors: Javier Alonso Villegas Luis, Marco Antonio Sobrevilla Cabezudo
Abstract: Linguistic features remain essential for interpretability and tasks involving style, structure, and readability, but existing Spanish tools offer limited coverage. We present PUCP-Metrix, an open-source repository of 182 linguistic metrics spanning lexical diversity, syntactic and semantic complexity, cohesion, psycholinguistics, and readability. PUCP-Metrix enables fine-grained, interpretable text analysis. We evaluate its usefulness on Automated Readability Assessment and Machine-Generated Text Detection, showing competitive performance compared to an existing repository and strong neural baselines. PUCP-Metrix offers a comprehensive, extensible resource for Spanish, supporting diverse NLP applications.
Authors: Yesheng Liu, Hao Li, Haiyu Xu, Baoqi Pei, Jiahao Wang, Mingxuan Zhao, Jingshu Zheng, Zheqi He, JG Yao, Bowen Qin, Xi Yang, Jiajun Zhang
Abstract: Multiple-choice question answering (MCQA) has been a popular format for evaluating and reinforcement fine-tuning (RFT) of modern multimodal language models. Its constrained output format allows for simplified, deterministic automatic verification. However, we find that the options may leak exploitable signals, which makes the accuracy metrics unreliable for indicating real capabilities and encourages explicit or implicit answer guessing behaviors during RFT. We propose ReVeL (Rewrite and Verify by LLM), a framework that rewrites multiple-choice questions into open-form questions while keeping answers verifiable whenever possible. The framework categorizes questions according to different answer types, apply different rewriting and verification schemes, respectively. When applied for RFT, we converted 20k MCQA examples and use GRPO to finetune Qwen2.5-VL models. Models trained on ReVeL-OpenQA match MCQA accuracy on multiple-choice benchmarks and improve OpenQA accuracy by about six percentage points, indicating better data efficiency and more robust reward signals than MCQA-based training. When used for evaluation, ReVeL also reveals up to 20 percentage points of score inflation in MCQA benchmarks (relative to OpenQA), improves judging accuracy, and reduces both cost and latency. We will release code and data publicly.
Authors: Shrikant Kendre, Austin Xu, Honglu Zhou, Michael Ryoo, Shafiq Joty, Juan Carlos Niebles
Abstract: Traditional evaluation metrics for textual and visual question answering, like ROUGE, METEOR, and Exact Match (EM), focus heavily on n-gram based lexical similarity, often missing the deeper semantic understanding needed for accurate assessment. While measures like BERTScore and MoverScore leverage contextual embeddings to address this limitation, they lack flexibility in balancing sentence-level and keyword-level semantics and ignore lexical similarity, which remains important. Large Language Model (LLM) based evaluators, though powerful, come with drawbacks like high costs, bias, inconsistency, and hallucinations. To address these issues, we introduce SMILE: Semantic Metric Integrating Lexical Exactness, a novel approach that combines sentence-level semantic understanding with keyword-level semantic understanding and easy keyword matching. This composite method balances lexical precision and semantic relevance, offering a comprehensive evaluation. Extensive benchmarks across text, image, and video QA tasks show SMILE is highly correlated with human judgments and computationally lightweight, bridging the gap between lexical and semantic evaluation.
Authors: Zhen Wang, Zhifeng Gao, Guolin Ke
Abstract: Test-time scaling has been shown to substantially improve large language models' (LLMs) mathematical reasoning. However, for a large portion of mathematical corpora, especially theorem proving, RLVR's scalability is limited: intermediate reasoning is crucial, while final answers are difficult to directly and reliably verify. Meanwhile, token-level SFT often degenerates into rote memorization rather than inducing longer chains of thought. Inspired by BERT's self-supervised tasks, we propose MR-RLVR (Masked-and-Reordered RLVR), which constructs process-level self-supervised rewards via "masked-then-fill" and "step reordering" to extract learnable signals from intermediate reasoning. Our training pipeline comprises two stages: we first perform self-supervised training on sampled mathematical calculation and proof data; we then conduct RLVR fine-tuning on mathematical calculation datasets where only outcomes are verifiable. We implement MR-RLVR on Qwen2.5-3B and DeepSeek-R1-Distill-Qwen-1.5B, and evaluate on AIME24, AIME25, AMC23, and MATH500. Under a fixed sampling and decoding budget, MR-RLVR achieves average relative gains over the original RLVR of +9.86% Pass@1, +5.27% Pass@5, and +4.00% Pass@8. These results indicate that incorporating process-aware self-supervised signals can effectively enhance RLVR's scalability and performance in only outcome-verifiable settings.
Authors: Thorsten Fr\"ohlich, Tim Schlippe
Abstract: The evaluation of academic theses is a cornerstone of higher education, ensuring rigor and integrity. Traditional methods, though effective, are time-consuming and subject to evaluator variability. This paper presents RubiSCoT, an AI-supported framework designed to enhance thesis evaluation from proposal to final submission. Using advanced natural language processing techniques, including large language models, retrieval-augmented generation, and structured chain-of-thought prompting, RubiSCoT offers a consistent, scalable solution. The framework includes preliminary assessments, multidimensional assessments, content extraction, rubric-based scoring, and detailed reporting. We present the design and implementation of RubiSCoT, discussing its potential to optimize academic assessment processes through consistent, scalable, and transparent evaluation.
Authors: Nikesh Gyawali, Doina Caragea, Cornelia Caragea, Saif M. Mohammad
Abstract: In this work, we study English-language vaccine discourse in social media posts, specifically posts on X (formerly Twitter), in seven years before the COVID-19 outbreak (2013 to 2019) and three years after the outbreak was first reported (2020 to 2022). Drawing on theories from social cognition and the stereotype content model in Social Psychology, we analyze how English speakers talk about vaccines on social media to understand the evolving narrative around vaccines in social media posts. To do that, we first introduce a novel dataset comprising 18.7 million curated posts on vaccine discourse from 2013 to 2022. This extensive collection-filtered down from an initial 129 million posts through rigorous preprocessing-captures both pre-COVID and COVID-19 periods, offering valuable insights into the evolution of English-speaking X users' perceptions related to vaccines. Our analysis shows that the COVID-19 pandemic led to complex shifts in X users' sentiment and discourse around vaccines. We observe that negative emotion word usage decreased during the pandemic, with notable rises in usage of surprise, and trust related emotion words. Furthermore, vaccine-related language tended to use more warmth-focused words associated with trustworthiness, along with positive, competence-focused words during the early days of the pandemic, with a marked rise in negative word usage towards the end of the pandemic, possibly reflecting a growing vaccine hesitancy and skepticism.
Authors: Oliver Kramer
Abstract: Cognitive BASIC is a minimal, BASIC-style prompting language and in-model interpreter that structures large language model (LLM) reasoning into explicit, stepwise execution traces. Inspired by the simplicity of retro BASIC, we repurpose numbered lines and simple commands as an interpretable cognitive control layer. Modern LLMs can reliably simulate such short programs, enabling transparent multi-step reasoning inside the model. A natural-language interpreter file specifies command semantics, memory updates, and logging behavior. Our mental-model interpreter extracts declarative and procedural knowledge, detects contradictions, and produces resolutions when necessary. A comparison across three LLMs on a benchmark of knowledge extraction, conflict detection, and reasoning tasks shows that all models can execute Cognitive BASIC programs, with overall strong but not uniform performance.
Authors: Sang Truong, Yuheng Tu, Michael Hardy, Anka Reuel, Zeyu Tang, Jirayu Burapacheep, Jonathan Perera, Chibuike Uwakwe, Ben Domingue, Nick Haber, Sanmi Koyejo
Abstract: Benchmarks are pivotal in driving AI progress, and invalid benchmark questions frequently undermine their reliability. Manually identifying and correcting errors among thousands of benchmark questions is not only infeasible but also a critical bottleneck for reliable evaluation. In this work, we introduce a framework for systematic benchmark revision that leverages statistical analysis of response patterns to flag potentially invalid questions for further expert review. Our approach builds on a core assumption commonly used in AI evaluations that the mean score sufficiently summarizes model performance. This implies a unidimensional latent construct underlying the measurement experiment, yielding expected ranges for various statistics for each item. When empirically estimated values for these statistics fall outside the expected range for an item, the item is more likely to be problematic. Across nine widely used benchmarks, our method guides expert review to identify problematic questions with up to 84\% precision. In addition, we introduce an LLM-judge first pass to review questions, further reducing human effort. Together, these components provide an efficient and scalable framework for systematic benchmark revision.
Authors: Chenyang Shao, Dehao Huang, Yu Li, Keyu Zhao, Weiquan Lin, Yining Zhang, Qingbin Zeng, Zhiyu Chen, Tianxing Li, Yifei Huang, Taozhong Wu, Xinyang Liu, Ruotong Zhao, Mengsheng Zhao, Xuhua Zhang, Yue Wang, Yuanyi Zhen, Fengli Xu, Yong Li, Tie-Yan Liu
Abstract: With the rapid development of Large Language Models (LLMs), AI agents have demonstrated increasing proficiency in scientific tasks, ranging from hypothesis generation and experimental design to manuscript writing. Such agent systems are commonly referred to as "AI Scientists." However, existing AI Scientists predominantly formulate scientific discovery as a standalone search or optimization problem, overlooking the fact that scientific research is inherently a social and collaborative endeavor. Real-world science relies on a complex scientific infrastructure composed of collaborative mechanisms, contribution attribution, peer review, and structured scientific knowledge networks. Due to the lack of modeling for these critical dimensions, current systems struggle to establish a genuine research ecosystem or interact deeply with the human scientific community. To bridge this gap, we introduce OmniScientist, a framework that explicitly encodes the underlying mechanisms of human research into the AI scientific workflow. OmniScientist not only achieves end-to-end automation across data foundation, literature review, research ideation, experiment automation, scientific writing, and peer review, but also provides comprehensive infrastructural support by simulating the human scientific system, comprising: (1) a structured knowledge system built upon citation networks and conceptual correlations; (2) a collaborative research protocol (OSP), which enables seamless multi-agent collaboration and human researcher participation; and (3) an open evaluation platform (ScienceArena) based on blind pairwise user voting and Elo rankings. This infrastructure empowers agents to not only comprehend and leverage human knowledge systems but also to collaborate and co-evolve, fostering a sustainable and scalable innovation ecosystem.
Authors: Patrick Amadeus Irawan, Ikhlasul Akmal Hanif, Muhammad Dehan Al Kautsar, Genta Indra Winata, Fajri Koto, Alham Fikri Aji
Abstract: Although the cultural dimension has been one of the key aspects in evaluating Vision-Language Models (VLMs), their ability to remain stable across diverse cultural inputs remains largely untested, despite being crucial to support diversity and multicultural societies. Existing evaluations often rely on benchmarks featuring only a singular cultural concept per image, overlooking scenarios where multiple, potentially unrelated cultural cues coexist. To address this gap, we introduce ConfusedTourist, a novel cultural adversarial robustness suite designed to assess VLMs' stability against perturbed geographical cues. Our experiments reveal a critical vulnerability, where accuracy drops heavily under simple image-stacking perturbations and even worsens with its image-generation-based variant. Interpretability analyses further show that these failures stem from systematic attention shifts toward distracting cues, diverting the model from its intended focus. These findings highlight a critical challenge: visual cultural concept mixing can substantially impair even state-of-the-art VLMs, underscoring the urgent need for more culturally robust multimodal understanding.
Authors: Arnau Mir, Alejandro Mus, Juan Vicente Riera
Abstract: Discrete fuzzy numbers, and in particular those defined over a finite chain $L_n = \{0, \ldots, n\}$, have been effectively employed to represent linguistic information within the framework of fuzzy systems. Research on total (admissible) orderings of such types of fuzzy subsets, and specifically those belonging to the set $\mathcal{D}_1^{L_n\rightarrow Y_m}$ consisting of discrete fuzzy numbers $A$ whose support is a closed subinterval of the finite chain $L_n = \{0, 1, \ldots, n\}$ and whose membership values $A(x)$, for $x \in L_n$, belong to the set $Y_m = \{ 0 = y_1 < y_2 < \cdots < y_{m-1} < y_m = 1 \}$, has facilitated the development of new methods for constructing logical connectives, based on a bijective function, called $\textit{pos function}$, that determines the position of each $A \in \mathcal{D}_1^{L_n\rightarrow Y_m}$. For this reason, in this work we revisit the problem by introducing algorithms that exploit the combinatorial structure of total (admissible) orders to compute the $\textit{pos}$ function and its inverse with exactness. The proposed approach achieves a complexity of $\mathcal{O}(n^{2} m \log n)$, which is quadratic in the size of the underlying chain ($n$) and linear in the number of membership levels ($m$). The key point is that the dominant factor is $m$, ensuring scalability with respect to the granularity of membership values. The results demonstrate that this formulation substantially reduces computational cost and enables the efficient implementation of algebraic operations -- such as aggregation and implication -- on the set of discrete fuzzy numbers.
Authors: Duo Zhou, Yuji Zhang, Tianxin Wei, Ruizhong Qiu, Ke Yang, Xiao Lin, Cheng Qian, Jingrui He, Hanghang Tong, Heng Ji, Huan Zhang
Abstract: Machine unlearning, the removal of a training subset's influence from a deployed model, is critical for privacy preservation and model reliability, yet gradient ascent on forget samples often harms retained knowledge. Existing approaches face a persistent tradeoff between effective forgetting and preservation on the retain set. While previous methods provide useful heuristics, they often lack a formal analysis on how exactly forgetting updates harm retained knowledge, and whether the side effects can be removed with theoretical guarantees. To explore a theoretically sound and simple solution, we start from the first principle on how performance on the retain set is actually affected: a first-order analysis of the local change of the retain loss under small parameter updates during model training. We start from a crisp equivalence: the retain loss is unchanged to first order iff the update direction is orthogonal to the subspace spanned by retain gradients ("retain-invariant"). This identifies the entangled component as the tangential part of forget update within the retain-gradient subspace, and characterizes disentanglement as orthogonality. Guided by this, we propose the Geometric-disentanglement Unlearning (GU) that decomposes any candidate forget gradient update into tangential and normal components to retain space and executes only the normal component. Under a standard trust-region budget, the projected direction aligned with the raw forget gradient is optimal among all first-order retain-invariant moves, and we also derive the optimal projected direction for joint forget-retain updating objectives. Our method is plug-and-play and can be attached to existing gradient-based unlearning procedures to mitigate side effects. GU achieves consistent improvement on various methods across three benchmarks TOFU, MUSE, and WMDP.
Authors: Haijiang Liu, Jinguang Gu, Xun Wu, Daniel Hershcovich, Qiaoling Xiao
Abstract: As Large Language Models (LLMs) increasingly influence high-stakes decision-making across global contexts, ensuring their alignment with diverse cultural values has become a critical governance challenge. This study presents a Multi-Layered Auditing Platform for Responsible AI that systematically evaluates cross-cultural value alignment in China-origin and Western-origin LLMs through four integrated methodologies: Ethical Dilemma Corpus for assessing temporal stability, Diversity-Enhanced Framework (DEF) for quantifying cultural fidelity, First-Token Probability Alignment for distributional accuracy, and Multi-stAge Reasoning frameworK (MARK) for interpretable decision-making. Our comparative analysis of 20+ leading models, such as Qwen, GPT-4o, Claude, LLaMA, and DeepSeek, reveals universal challenges-fundamental instability in value systems, systematic under-representation of younger demographics, and non-linear relationships between model scale and alignment quality-alongside divergent regional development trajectories. While China-origin models increasingly emphasize multilingual data integration for context-specific optimization, Western models demonstrate greater architectural experimentation but persistent U.S.-centric biases. Neither paradigm achieves robust cross-cultural generalization. We establish that Mistral-series architectures significantly outperform LLaMA3-series in cross-cultural alignment, and that Full-Parameter Fine-Tuning on diverse datasets surpasses Reinforcement Learning from Human Feedback in preserving cultural variation...
Authors: Callie C. Liao, Duoduo Liao, Ellie L. Zhang
Abstract: Recent advances in generative AI have made music generation a prominent research focus. However, many neural-based models rely on large datasets, raising concerns about copyright infringement and high-performance costs. In contrast, we propose MusicAIR, an innovative multimodal AI music generation framework powered by a novel algorithm-driven symbolic music core, effectively mitigating copyright infringement risks. The music core algorithms connect critical lyrical and rhythmic information to automatically derive musical features, creating a complete, coherent melodic score solely from the lyrics. The MusicAIR framework facilitates music generation from lyrics, text, and images. The generated score adheres to established principles of music theory, lyrical structure, and rhythmic conventions. We developed Generate AI Music (GenAIM), a web tool using MusicAIR for lyric-to-song, text-to-music, and image-to-music generation. In our experiments, we evaluated AI-generated music scores produced by the system using both standard music metrics and innovative analysis that compares these compositions with original works. The system achieves an average key confidence of 85%, outperforming human composers at 79%, and aligns closely with established music theory standards, demonstrating its ability to generate diverse, human-like compositions. As a co-pilot tool, GenAIM can serve as a reliable music composition assistant and a possible educational composition tutor while simultaneously lowering the entry barrier for all aspiring musicians, which is innovative and significantly contributes to AI for music generation.
Authors: Chiori Hori, Yoshiki Masuyama, Siddarth Jain, Radu Corcodel, Devesh Jha, Diego Romeres, Jonathan Le Roux
Abstract: Human-robot collaboration towards a shared goal requires robots to understand human action and interaction with the surrounding environment. This paper focuses on human-robot interaction (HRI) based on human-robot dialogue that relies on the robot action confirmation and action step generation using multimodal scene understanding. The state-of-the-art approach uses multimodal transformers to generate robot action steps aligned with robot action confirmation from a single clip showing a task composed of multiple micro steps. Although actions towards a long-horizon task depend on each other throughout an entire video, the current approaches mainly focus on clip-level processing and do not leverage long-context information. This paper proposes a long-context Q-former incorporating left and right context dependency in full videos. Furthermore, this paper proposes a text-conditioning approach to feed text embeddings directly into the LLM decoder to mitigate the high abstraction of the information in text by Q-former. Experiments with the YouCook2 corpus show that the accuracy of confirmation generation is a major factor in the performance of action planning. Furthermore, we demonstrate that the long-context Q-former improves the confirmation and action planning by integrating VideoLLaMA3.
Authors: Yidong Huang, Zun Wang, Han Lin, Dong-Ki Kim, Shayegan Omidshafiei, Jaehong Yoon, Yue Zhang, Mohit Bansal
Abstract: Recent video generation approaches increasingly rely on planning intermediate control signals such as object trajectories to improve temporal coherence and motion fidelity. However, these methods mostly employ single-shot plans that are typically limited to simple motions, or iterative refinement which requires multiple calls to the video generator, incuring high computational cost. To overcome these limitations, we propose SketchVerify, a training-free, sketch-verification-based planning framework that improves motion planning quality with more dynamically coherent trajectories (i.e., physically plausible and instruction-consistent motions) prior to full video generation by introducing a test-time sampling and verification loop. Given a prompt and a reference image, our method predicts multiple candidate motion plans and ranks them using a vision-language verifier that jointly evaluates semantic alignment with the instruction and physical plausibility. To efficiently score candidate motion plans, we render each trajectory as a lightweight video sketch by compositing objects over a static background, which bypasses the need for expensive, repeated diffusion-based synthesis while achieving comparable performance. We iteratively refine the motion plan until a satisfactory one is identified, which is then passed to the trajectory-conditioned generator for final synthesis. Experiments on WorldModelBench and PhyWorldBench demonstrate that our method significantly improves motion quality, physical realism, and long-term consistency compared to competitive baselines while being substantially more efficient. Our ablation study further shows that scaling up the number of trajectory candidates consistently enhances overall performance.
Authors: Yuxian Gu, Li Dong, Furu Wei, Minlie Huang
Abstract: Knowledge Distillation (KD) is a promising technique for reducing the high computational demand of large language models (LLMs). However, previous KD methods are primarily applied to white-box classification models or training small models to imitate black-box model APIs like ChatGPT. How to effectively distill the knowledge of white-box LLMs into small models is still under-explored, which becomes more important with the prosperity of open-source LLMs. In this work, we propose a KD approach that distills LLMs into smaller language models. We first replace the forward Kullback-Leibler divergence (KLD) objective in the standard KD approaches with reverse KLD, which is more suitable for KD on generative language models, to prevent the student model from overestimating the low-probability regions of the teacher distribution. Then, we derive an effective on-policy optimization approach to learn this objective. The student models are named MiniLLM. Extensive experiments in the instruction-following setting show that MiniLLM generates more precise responses with higher overall quality, lower exposure bias, better calibration, and higher long-text generation performance than the baselines. Our method is scalable for different model families with 120M to 13B parameters. Our code, data, and model checkpoints can be found in https://github.com/microsoft/LMOps/tree/main/minillm.
Authors: Miguel Moura Ramos, Tom\'as Almeida, Daniel Vareta, Filipe Azevedo, Sweta Agrawal, Patrick Fernandes, Andr\'e F. T. Martins
Abstract: Reinforcement learning (RL) has been proven to be an effective and robust method for training neural machine translation systems, especially when paired with powerful reward models that accurately assess translation quality. However, most research has focused on RL methods that use sentence-level feedback, leading to inefficient learning signals due to the reward sparsity problem -- the model receives a single score for the entire sentence. To address this, we propose a novel approach that leverages fine-grained, token-level quality assessments along with error severity levels using RL methods. Specifically, we use xCOMET, a state-of-the-art quality estimation system, as our token-level reward model. We conduct experiments on small and large translation datasets with standard encoder-decoder and large language models-based machine translation systems, comparing the impact of sentence-level versus fine-grained reward signals on translation quality. Our results show that training with token-level rewards improves translation quality across language pairs over baselines according to both automatic and human evaluation. Furthermore, token-level reward optimization improves training stability, evidenced by a steady increase in mean rewards over training epochs.
Authors: Hang Gao, Yongfeng Zhang
Abstract: By augmenting Large Language Models (LLMs) with external tools, their capacity to solve complex problems has been significantly enhanced. However, despite ongoing advancements in the parsing capabilities of LLMs, incorporating all available tools simultaneously in the prompt remains impractical due to the vast number of external tools. Consequently, it is essential to provide LLMs with a precise set of tools tailored to the specific task, considering both quantity and quality. Current tool retrieval methods primarily focus on refining the ranking list of tools and directly packaging a fixed number of top-ranked tools as the tool set. However, these approaches often fail to equip LLMs with the optimal set of tools prior to execution, since the optimal number of tools for different tasks could be different, resulting in inefficiencies such as redundant or unsuitable tools, which impede immediate access to the most relevant tools. This paper addresses the challenge of recommending precise toolsets for LLMs. We introduce the problem of tool recommendation, define its scope, and propose a novel Precision-driven Tool Recommendation (PTR) approach. PTR captures an initial, concise set of tools by leveraging historical tool bundle usage and dynamically adjusts the tool set by performing tool matching, culminating in a multi-view-based tool addition. Additionally, we present a new dataset, RecTools, and a metric, TRACC, designed to evaluate the effectiveness of tool recommendation for LLMs. We further validate our design choices through comprehensive experiments, demonstrating promising accuracy across two open benchmarks and our RecTools dataset.
Authors: Zhengyi Zhao, Shubo Zhang, Yiming Du, Bin Liang, Baojun Wang, Zhongyang Li, Binyang Li, Kam-Fai Wong
Abstract: Large language models have improved dialogue systems, but often process conversational turns in isolation, overlooking the event structures that guide natural interactions. Hence we introduce \textbf{EventWeave}, a framework that explicitly models relationships between conversational events to generate more contextually appropriate dialogue responses. EventWeave constructs a dynamic event graph that distinguishes between core events (main goals) and supporting events (interconnected details), employing a multi-head attention mechanism to selectively determine which events are most relevant to the current turn. Unlike summarization or standard graph-based approaches, our method captures three distinct relationship types between events, allowing for more nuanced context modeling. Experiments on three dialogue datasets demonstrate that EventWeave produces more natural and contextually appropriate responses while requiring less computational overhead than models processing the entire dialogue history. Ablation studies confirm improvements stem from better event relationship modeling rather than increased information density. Our approach effectively balances comprehensive context understanding with generating concise responses, maintaining strong performance across various dialogue lengths through targeted optimization techniques.
Authors: Mehdi Fatemi, Banafsheh Rafiee, Mingjie Tang, Kartik Talamadupula
Abstract: A major drawback of reasoning models is their excessive token usage, inflating computational cost, resource demand, and latency. We show this verbosity stems not from deeper reasoning but from reinforcement learning loss minimization when models produce incorrect answers. With unsolvable problems dominating training, this effect compounds into a systematic tendency toward longer outputs. Through theoretical analysis of PPO and GRPO, we prove that incorrect answers inherently drive policies toward verbosity \textit{even when} $\gamma=1$, reframing response lengthening as an optimization artifact. We further uncover a consistent correlation between conciseness and correctness across reasoning and non-reasoning models. Building on these insights, we propose a two-phase RL procedure where a brief secondary stage, trained on a small set of solvable problems, significantly reduces response length while preserving or improving accuracy. Finally, we show that while GRPO shares properties with PPO, it exhibits collapse modes, limiting its reliability for concise reasoning. Our claims are supported by extensive experiments.
Authors: Yihuai Hong, Yiran Zhao, Wei Tang, Yang Deng, Yu Rong, Wenxuan Zhang
Abstract: Over time, a growing wave of large language models from various series has been introduced to the community. Researchers are striving to maximize the performance of language models with constrained parameter sizes. However, from a microscopic perspective, there has been limited research on how to better store knowledge in model parameters, particularly within MLPs, to enable more effective utilization of this knowledge by the model. In this work, we analyze twenty publicly available open-source large language models to investigate the relationship between their strong performance and the way knowledge is stored in their corresponding MLP parameters. Our findings reveal that as language models become more advanced and demonstrate stronger knowledge capabilities, their parameters exhibit increased specialization. Specifically, parameters in the MLPs tend to be more focused on encoding similar types of knowledge. We experimentally validate that this specialized distribution of knowledge contributes to improving the efficiency of knowledge utilization in these models. Furthermore, by conducting causal training experiments, we confirm that this specialized knowledge distribution plays a critical role in improving the model's efficiency in leveraging stored knowledge.
Authors: Beong-woo Kwak, Minju Kim, Dongha Lim, Hyungjoo Chae, Dongjin Kang, Sunghwan Kim, Dongil Yang, Jinyoung Yeo
Abstract: Large language models (LLMs) have demonstrated strong capabilities in using external tools to address user inquiries. However, most existing evaluations assume tool use in short contexts, offering limited insight into model behavior during realistic long-term interactions. To fill this gap, we introduce ToolHaystack, a benchmark for testing the tool use capabilities in long-term interactions. Each test instance in ToolHaystack includes multiple tasks execution contexts and realistic noise within a continuous conversation, enabling assessment of how well models maintain context and handle various disruptions. By applying this benchmark to 14 state-of-the-art LLMs, we find that while current models perform well in standard multi-turn settings, they often significantly struggle in ToolHaystack, highlighting critical gaps in their long-term robustness not revealed by previous tool benchmarks.
Authors: Haining Wang, Jason Clark, Yueru Yan, Star Bradley, Ruiyang Chen, Yiqiong Zhang, Hengyi Fu, Zuoyu Tian
Abstract: As libraries explore large language models (LLMs) for use in virtual reference services, a key question arises: Can LLMs serve all users equitably, regardless of demographics or social status? While they offer great potential for scalable support, LLMs may also reproduce societal biases embedded in their training data, risking the integrity of libraries' commitment to equitable service. To address this concern, we evaluate whether LLMs differentiate responses across user identities by prompting six state-of-the-art LLMs to assist patrons differing in sex, race/ethnicity, and institutional role. We find no evidence of differentiation by race or ethnicity, and only minor evidence of stereotypical bias against women in one model. LLMs demonstrate nuanced accommodation of institutional roles through the use of linguistic choices related to formality, politeness, and domain-specific vocabularies, reflecting professional norms rather than discriminatory treatment. These findings suggest that current LLMs show a promising degree of readiness to support equitable and contextually appropriate communication in academic library reference services.
Authors: Ziqi Miao, Lijun Li, Yuan Xiong, Zhenhua Liu, Pengyu Zhu, Jing Shao
Abstract: Contextual priming, where earlier stimuli covertly bias later judgments, offers an unexplored attack surface for large language models (LLMs). We uncover a contextual priming vulnerability in which the previous response in the dialogue can steer its subsequent behavior toward policy-violating content. While existing jailbreak attacks largely rely on single-turn or multi-turn prompt manipulations, or inject static in-context examples, these methods suffer from limited effectiveness, inefficiency, or semantic drift. We introduce Response Attack (RA), a novel framework that strategically leverages intermediate, mildly harmful responses as contextual primers within a dialogue. By reformulating harmful queries and injecting these intermediate responses before issuing a targeted trigger prompt, RA exploits a previously overlooked vulnerability in LLMs. Extensive experiments across eight state-of-the-art LLMs show that RA consistently achieves significantly higher attack success rates than nine leading jailbreak baselines. Our results demonstrate that the success of RA is directly attributable to the strategic use of intermediate responses, which induce models to generate more explicit and relevant harmful content while maintaining stealth, efficiency, and fidelity to the original query. The code and data are available at https://github.com/Dtc7w3PQ/Response-Attack.
Authors: David M. Markowitz, Samuel Hardman Taylor
Abstract: We examined how online hate is motivated by receiving social approval via Walther's (2024) social approval theory of online hate, which argues (H1a) more signals of social approval on hate messages predicts more subsequent hate messages, and (H1b) as social approval increases, hate speech becomes more extreme. Using 110 million messages from Parler (2018-2021), we observed the number of upvotes received on a hate speech post was unassociated with hate speech in one's next post and during the next month, three-months, and six-months. The number of upvotes received on (extreme) hate speech comments, however, was positively associated with (extreme) hate speech during the next week, month, three-months, and six-months. Between-person effects revealed an average positive relationship between social approval and hate speech production at all time intervals. For comments, social approval linked more strongly to online hate than social disapproval. Social approval is a critical mechanism facilitating online hate propagation.
Authors: Kelly Kendro, Jeffrey Maloney, Scott Jarvis
Abstract: The degree to which large language models (LLMs) produce writing that is truly human-like remains unclear despite the extensive empirical attention that this question has received. The present study addresses this question from the perspective of lexical diversity. Specifically, the study investigates patterns of lexical diversity in LLM-generated texts from four ChatGPT models (ChatGPT-3.5, ChatGPT-4, ChatGPT-o4 mini, and ChatGPT-4.5) in comparison with texts written by L1 and L2 English participants (n = 240) across four education levels. Six dimensions of lexical diversity were measured in each text: volume, abundance, variety-repetition, evenness, disparity, and dispersion. Results from one-way MANOVAs, one-way ANOVAs, and Support Vector Machines revealed that the ChatGPT-generated texts differed significantly from human-written texts for each variable, with ChatGPT-o4 mini and ChatGPT-4.5 differing the most. Within these two groups, ChatGPT-4.5 demonstrated higher levels of lexical diversity than older models despite producing fewer tokens. The human writers' lexical diversity did not differ across subgroups (i.e., education, language status). Altogether, the results indicate that ChatGPT models do not produce human-like texts in relation to lexical diversity, and the newer models produce less human-like text than older models. We discuss the implications of these results for language pedagogy and related applications.
Authors: Maciej Skorski, Alina Landowska
Abstract: How do Large Language Models understand moral dimensions compared to humans? This first large-scale Bayesian evaluation of market-leading language models provides the answer. In contrast to prior work using deterministic ground truth (majority or inclusion rules), we model annotator disagreements to capture both aleatoric uncertainty (inherent human disagreement) and epistemic uncertainty (model domain sensitivity). We evaluated the best language models (Claude Sonnet 4, DeepSeek-V3, Llama 4 Maverick) across 250K+ annotations from nearly 700 annotators in 100K+ texts spanning social networks, news and forums. Our GPU-optimized Bayesian framework processed 1M+ model queries, revealing that AI models typically rank among the top 25\% of human annotators, performing much better than average balanced accuracy. Importantly, we find that AI produces far fewer false negatives than humans, highlighting their more sensitive moral detection capabilities.
Authors: Chia-Hsuan Hsu, Jun-En Ding, Hsin-Ling Hsu, Chih-Ho Hsu, Li-Hung Yao, Chun-Chieh Liao, Feng Liu, Fang-Ming Hung
Abstract: Medical question answering requires advanced reasoning that integrates domain knowledge with logical inference. However, existing large language models (LLMs) often generate reasoning chains that lack factual accuracy and clinical reliability. We propose Ranked Preference Reinforcement Optimization (RPRO), a novel framework that combines reinforcement learning with preference-driven reasoning refinement to enhance clinical chain-of-thought (CoT) performance. RPRO distinguishes itself from prior approaches by employing task-adaptive reasoning templates and a probabilistic evaluation mechanism that aligns model outputs with established clinical workflows, while automatically identifying and correcting low-quality reasoning chains. Unlike traditional pairwise preference methods, RPRO introduces a groupwise ranking optimization based on the Bradley--Terry model and incorporates KL-divergence regularization for stable training. Experiments on PubMedQA, MedQA-USMLE, and a real-world clinical dataset from Far Eastern Memorial Hospital (FEMH) demonstrate consistent improvements over strong baselines. Remarkably, our 2B-parameter model outperforms much larger 7B--20B models, including medical-specialized variants. These findings demonstrate that combining preference optimization with quality-driven refinement provides a scalable and clinically grounded approach to building more reliable medical LLMs.
Authors: Lovely Yeswanth Panchumarthi, Sai Prasad Gudari, Atharva Negi, Praveen Raj Budime, Harsit Upadhya
Abstract: The exponential growth of biomedical literature creates significant challenges for accessing precise medical information. Current biomedical question-answering systems primarily focus on short-form answers, failing to provide the comprehensive explanations necessary for clinical decision-making. We present RAG-BioQA, a novel framework combining retrieval-augmented generation with domain-specific fine-tuning to produce evidence-based, long-form biomedical answers. Our approach integrates BioBERT embeddings with FAISS indexing and compares various re-ranking strategies (BM25, ColBERT, MonoT5) to optimize context selection before synthesizing evidence through a fine-tuned T5 model. Experimental results on the PubMedQA dataset show significant improvements over baselines, with our best model achieving substantial gains across BLEU, ROUGE, and METEOR metrics, advancing the state of accessible, evidence-based biomedical knowledge retrieval.
Authors: Pablo Miralles-Gonz\'alez, Javier Huertas-Tato, Alejandro Mart\'in, David Camacho
Abstract: Computational stylometry analyzes writing style through quantitative patterns in text, supporting applications from forensic tasks such as identity linking and plagiarism detection to literary attribution in the humanities. Supervised and contrastive approaches rely on data with spurious correlations and often confuse style with topic. Despite their natural use in AI-generated text detection, the CLM pre-training of modern LLMs has been scarcely leveraged for general authorship problems. We propose a novel unsupervised approach based on this extensive pre-training and the in-context learning capabilities of LLMs, employing the log-probabilities of an LLM to measure style transferability from one text to another. Our method significantly outperforms LLM prompting approaches of comparable scale and achieves higher accuracy than contrastively trained baselines when controlling for topical correlations. Moreover, performance scales fairly consistently with the size of the base model and, in the case of authorship verification, with an additional mechanism that increases test-time computation; enabling flexible trade-offs between computational cost and accuracy.
Authors: Ashish Kattamuri, Ishita Prasad, Meetu Malhotra, Arpita Vats, Rahul Raja, Albert Lie
Abstract: Current Text-to-SQL methods are evaluated and only focused on executable queries, overlooking the semantic alignment challenge -- both in terms of the semantic meaning of the query and the correctness of the execution results. Even execution accuracy itself shows significant drops when moving from English to other languages, with an average decline of 6 percentage points across non-English languages. We address these challenges by presenting a new framework that combines Group Relative Policy Optimization (GRPO) within a multilingual contrastive reward signal to enhance both task efficiency and semantic accuracy in Text-to-SQL systems in cross-lingual scenarios. Our method teaches models to obtain better correspondence between SQL generation and user intent by combining a reward signal based on semantic similarity. On the seven-language MultiSpider dataset, fine-tuning the LLaMA-3-3B model with GRPO improved the execution accuracy up to 87.4 percent (+26 pp over zero-shot) and semantic accuracy up to 52.29 percent (+32.86 pp). Adding our contrastive reward signal in the GRPO framework further improved the average semantic accuracy to 59.14 percent (+6.85 pp, up to +10 pp for Vietnamese). Our experiments showcase that a smaller, parameter-efficient 3B LLaMA model fine-tuned with our contrastive reward signal outperforms a much larger zero-shot 8B LLaMA model, with an uplift of 7.43 pp in execution accuracy (from 81.43 percent on the 8B model to 88.86 percent on the 3B model), and nearly matches its semantic accuracy (59.14 percent vs. 68.57 percent) -- all using just 3,000 reinforcement learning training examples. These results demonstrate how we can improve the performance of Text-to-SQL systems with contrastive rewards for directed semantic alignment, without requiring large-scale training datasets.
Authors: Haoxuan Zhang, Ruochi Li, Sarthak Shrestha, Shree Harshini Mamidala, Revanth Putta, Arka Krishan Aggarwal, Ting Xiao, Junhua Ding, Haihua Chen
Abstract: Peer review serves as the gatekeeper of science, yet the surge in submissions and widespread adoption of large language models (LLMs) in scholarly evaluation present unprecedented challenges. While recent work has focused on using LLMs to improve review efficiency, unchecked deficient reviews from both human experts and AI systems threaten to systematically undermine academic integrity. To address this issue, we introduce ReviewGuard, an automated system for detecting and categorizing deficient reviews through a four-stage LLM-driven framework: data collection from ICLR and NeurIPS on OpenReview, GPT-4.1 annotation with human validation, synthetic data augmentation yielding 6,634 papers with 24,657 real and 46,438 synthetic reviews, and fine-tuning of encoder-based models and open-source LLMs. Feature analysis reveals that deficient reviews exhibit lower rating scores, higher self-reported confidence, reduced structural complexity, and more negative sentiment than sufficient reviews. AI-generated text detection shows dramatic increases in AI-authored reviews since ChatGPT's emergence. Mixed training with synthetic and real data substantially improves detection performance - for example, Qwen 3-8B achieves recall of 0.6653 and F1 of 0.7073, up from 0.5499 and 0.5606 respectively. This study presents the first LLM-driven system for detecting deficient peer reviews, providing evidence to inform AI governance in peer review. Code, prompts, and data are available at https://github.com/haoxuan-unt2024/ReviewGuard
Authors: Jenna Russell, Marzena Karpinska, Destiny Akinode, Katherine Thai, Bradley Emi, Max Spero, Mohit Iyyer
Abstract: AI is rapidly transforming journalism, but the extent of its use in published newspaper articles remains unclear. We address this gap by auditing a large-scale dataset of 186K articles from online editions of 1.5K American newspapers published in the summer of 2025. Using Pangram, a state-of-the-art AI detector, we discover that approximately 9% of newly-published articles are either partially or fully AI-generated. This AI use is unevenly distributed, appearing more frequently in smaller, local outlets, in specific topics such as weather and technology, and within certain ownership groups. We also analyze 45K opinion pieces from Washington Post, New York Times, and Wall Street Journal, finding that they are 6.4 times more likely to contain AI-generated content than news articles from the same publications, with many AI-flagged op-eds authored by prominent public figures. Despite this prevalence, we find that AI use is rarely disclosed: a manual audit of 100 AI-flagged articles found only five disclosures of AI use. Overall, our audit highlights the immediate need for greater transparency and updated editorial standards regarding the use of AI in journalism to maintain public trust.
Authors: Mo El-Haj, Paul Rayson
Abstract: This paper examines how domain specificity affects abstractive summarisation of Arabic financial texts using large language models (LLMs). We present AraFinNews, the largest publicly available Arabic financial news dataset to date, comprising 212,500 article-headline pairs spanning almost a decade of reporting from October 2015 to July 2025. Developed as an Arabic counterpart to major English summarisation corpora such as CNN/DailyMail, AraFinNews offers a strong benchmark for assessing domain-focused language understanding and generation in financial contexts. Using this resource, we evaluate transformer-based models, including mT5, AraT5 and the domain-adapted FinAraT5, to investigate how financial-domain pretraining influences accuracy, numerical reliability and stylistic alignment with professional reporting. The results show that domain-adapted models produce more coherent summaries, particularly when handling quantitative and entity-centred information. These findings underscore the value of domain-specific adaptation for improving narrative fluency in Arabic financial summarisation. The dataset is freely available for non-commercial research at https://github.com/ArabicNLP-UK/AraFinNews.
Authors: C\'elian Ringwald, Fabien Gandon, Catherine Faron, Franck Michel, Hanna Abi Akl
Abstract: Small language models (SLMs) have shown promises for relation extraction (RE) when extracting RDF triples guided by SHACL shapes focused on common datatype properties. This paper investigates how SLMs handle both datatype and object properties for a complete RDF graph extraction. We show that the key bottleneck is related to long-tail distribution of rare properties. To solve this issue, we evaluate several strategies: stratified sampling, weighted loss, dataset scaling, and template-based synthetic data augmentation. We show that the best strategy to perform equally well over unbalanced target properties is to build a training set where the number of occurrences of each property exceeds a given threshold. To enable reproducibility, we publicly released our datasets, experimental results and code. Our findings offer practical guidance for training shape-aware SLMs and highlight promising directions for future work in semantic RE.
Authors: Ringwald Celian, Gandon, Fabien, Faron Catherine, Michel Franck, Abi Akl Hanna
Abstract: This article presents a systematic review of relation extraction (RE) research since the advent of Transformer-based models. Using an automated framework to collect and annotate publications, we analyze 34 surveys, 64 datasets, and 104 models published between 2019 and 2024. The review highlights methodological advances, benchmark resources, and the integration of semantic web technologies. By consolidating results across multiple dimensions, the study identifies current trends, limitations, and open challenges, offering researchers and practitioners a comprehensive reference for understanding the evolution and future directions of RE.
Authors: Eva Prakash, Maayane Attias, Pierre Chambon, Justin Xu, Steven Truong, Jean-Benoit Delbrouck, Tessa Cook, Curtis Langlotz
Abstract: Objective: To enhance automated de-identification of radiology reports by scaling transformer-based models through extensive training datasets and benchmarking performance against commercial cloud vendor systems for protected health information (PHI) detection. Materials and Methods: In this retrospective study, we built upon a state-of-the-art, transformer-based, PHI de-identification pipeline by fine-tuning on two large annotated radiology corpora from Stanford University, encompassing chest X-ray, chest CT, abdomen/pelvis CT, and brain MR reports and introducing an additional PHI category (AGE) into the architecture. Model performance was evaluated on test sets from Stanford and the University of Pennsylvania (Penn) for token-level PHI detection. We further assessed (1) the stability of synthetic PHI generation using a "hide-in-plain-sight" method and (2) performance against commercial systems. Precision, recall, and F1 scores were computed across all PHI categories. Results: Our model achieved overall F1 scores of 0.973 on the Penn dataset and 0.996 on the Stanford dataset, outperforming or maintaining the previous state-of-the-art model performance. Synthetic PHI evaluation showed consistent detectability (overall F1: 0.959 [0.958-0.960]) across 50 independently de-identified Penn datasets. Our model outperformed all vendor systems on synthetic Penn reports (overall F1: 0.960 vs. 0.632-0.754). Discussion: Large-scale, multimodal training improved cross-institutional generalization and robustness. Synthetic PHI generation preserved data utility while ensuring privacy. Conclusion: A transformer-based de-identification model trained on diverse radiology datasets outperforms prior academic and commercial systems in PHI detection and establishes a new benchmark for secure clinical text processing.
Authors: Shaowen Wang, Yiqi Dong, Ruinian Chang, Tansheng Zhu, Yuebo Sun, Kaifeng Lyu, Jian Li
Abstract: Despite substantial advances, large language models (LLMs) continue to exhibit hallucinations, generating plausible yet incorrect responses. In this paper, we highlight a critical yet previously underexplored class of hallucinations driven by spurious correlations -- superficial but statistically prominent associations between features (e.g., surnames) and attributes (e.g., nationality) present in the training data. We demonstrate that these spurious correlations induce hallucinations that are confidently generated, immune to model scaling, evade current detection methods, and persist even after refusal fine-tuning. Through systematically controlled synthetic experiments and empirical evaluations on state-of-the-art open-source and proprietary LLMs (including GPT-5), we show that existing hallucination detection methods, such as confidence-based filtering and inner-state probing, fundamentally fail in the presence of spurious correlations. Our theoretical analysis further elucidates why these statistical biases intrinsically undermine confidence-based detection techniques. Our findings thus emphasize the urgent need for new approaches explicitly designed to address hallucinations caused by spurious correlations.
Authors: Wenxin Zhu, Andong Chen, Yuchen Song, Kehai Chen, Conghui Zhu, Ziyan Chen, Tiejun Zhao
Abstract: With the remarkable success of Multimodal Large Language Models (MLLMs) in perception tasks, enhancing their complex reasoning capabilities has emerged as a critical research focus. Existing models still suffer from challenges such as opaque reasoning paths and insufficient generalization ability. Chain-of-Thought (CoT) reasoning, which has demonstrated significant efficacy in language models by enhancing reasoning transparency and output interpretability, holds promise for improving model reasoning capabilities when extended to the multimodal domain. This paper provides a systematic review centered on "Multimodal Chain-of-Thought" (MCoT). First, it analyzes the background and theoretical motivations for its inception from the perspectives of technical evolution and task demands. Then, it introduces mainstream MCoT methods from three aspects: CoT paradigms, the post-training stage, and the inference stage, while also analyzing their underlying mechanisms. Furthermore, the paper summarizes existing evaluation benchmarks and metrics, and discusses the application scenarios of MCoT. Finally, it analyzes the challenges currently facing MCoT and provides an outlook on its future research directions.
Authors: Mihai Nadas, Laura Diosan
Abstract: Automatic diacritic restoration is crucial for text processing in languages with rich diacritical marks, such as Romanian. This study evaluates the performance of several large language models (LLMs) in restoring diacritics in Romanian texts. Using a comprehensive corpus, we tested models including OpenAI's GPT-3.5, GPT-4, GPT-4o, Google's Gemini 1.0 Pro, Meta's Llama 2 and Llama 3, MistralAI's Mixtral 8x7B Instruct, airoboros 70B, and OpenLLM-Ro's RoLlama 2 7B, under multiple prompt templates ranging from zero-shot to complex multi-shot instructions. Results show that models such as GPT-4o achieve high diacritic restoration accuracy, consistently surpassing a neutral echo baseline, while others, including Meta's Llama family, exhibit wider variability. These findings highlight the impact of model architecture, training data, and prompt design on diacritic restoration performance and outline promising directions for improving NLP tools for diacritic-rich languages.
Authors: Zachary Ellis, Jared Joselowitz, Yash Deo, Yajie He, Anna Kalygina, Aisling Higham, Mana Rahimzadeh, Yan Jia, Ibrahim Habli, Ernest Lim
Abstract: As Automatic Speech Recognition (ASR) is increasingly deployed in clinical dialogue, standard evaluations still rely heavily on Word Error Rate (WER). This paper challenges that standard, investigating whether WER or other common metrics correlate with the clinical impact of transcription errors. We establish a gold-standard benchmark by having expert clinicians compare ground-truth utterances to their ASR-generated counterparts, labeling the clinical impact of any discrepancies found in two distinct doctor-patient dialogue datasets. Our analysis reveals that WER and a comprehensive suite of existing metrics correlate poorly with the clinician-assigned risk labels (No, Minimal, or Significant Impact). To bridge this evaluation gap, we introduce an LLM-as-a-Judge, programmatically optimized using GEPA through DSPy to replicate expert clinical assessment. The optimized judge (Gemini-2.5-Pro) achieves human-comparable performance, obtaining 90% accuracy and a strong Cohen's $\kappa$ of 0.816. This work provides a validated, automated framework for moving ASR evaluation beyond simple textual fidelity to a necessary, scalable assessment of safety in clinical dialogue.
Authors: Daiqing Wu, Dongbao Yang, Huawen Shen, Can Ma, Yu Zhou
Abstract: With the proliferation of social media posts in recent years, the need to detect sentiments in multimodal (image-text) content has grown rapidly. Since posts are user-generated, the image and text from the same post can express different or even contradictory sentiments, leading to potential \textbf{sentiment discrepancy}. However, existing works mainly adopt a single-branch fusion structure that primarily captures the consistent sentiment between image and text. The ignorance or implicit modeling of discrepant sentiment results in compromised unimodal encoding and limited performance. In this paper, we propose a semantics Completion and Decomposition (CoDe) network to resolve the above issue. In the semantics completion module, we complement image and text representations with the semantics of the in-image text, helping bridge the sentiment gap. In the semantics decomposition module, we decompose image and text representations with exclusive projection and contrastive learning, thereby explicitly capturing the discrepant sentiment between modalities. Finally, we fuse image and text representations by cross-attention and combine them with the learned discrepant sentiment for final classification. Extensive experiments on four datasets demonstrate the superiority of CoDe and the effectiveness of each proposed module.
Authors: Zekun Zhou, Xiaocheng Feng, Lei Huang, Xiachong Feng, Ziyun Song, Ruihan Chen, Liang Zhao, Weitao Ma, Yuxuan Gu, Baoxin Wang, Dayong Wu, Guoping Hu, Ting Liu, Bing Qin
Abstract: Research is a fundamental process driving the advancement of human civilization, yet it demands substantial time and effort from researchers. In recent years, the rapid development of artificial intelligence (AI) technologies has inspired researchers to explore how AI can accelerate and enhance research. To monitor relevant advancements, this paper presents a systematic review of the progress in this domain. Specifically, we organize the relevant studies into three main categories: hypothesis formulation, hypothesis validation, and manuscript publication. Hypothesis formulation involves knowledge synthesis and hypothesis generation. Hypothesis validation includes the verification of scientific claims, theorem proving, and experiment validation. Manuscript publication encompasses manuscript writing and the peer review process. Furthermore, we identify and discuss the current challenges faced in these areas, as well as potential future directions for research. Finally, we also offer a comprehensive overview of existing benchmarks and tools across various domains that support the integration of AI into the research process. We hope this paper serves as an introduction for beginners and fosters future research. Resources have been made publicly available at https://github.com/zkzhou126/AI-for-Research.
Authors: Soo Yong Lee, Hyunjin Hwang, Taekwan Kim, Yuyeong Kim, Kyuri Park, Jaemin Yoo, Denny Borsboom, Kijung Shin
Abstract: Can large language models (LLMs) instantiate computations of psychopathology? An effective approach to the question hinges on addressing two factors. First, for conceptual validity, we require a general and computational account of psychopathology that is applicable to computational entities without biological embodiment or subjective experience. Second, psychopathological computations, derived from the adapted theory, need to be empirically identified within the LLM's internal processing. Thus, we establish a computational-theoretical framework to provide an account of psychopathology applicable to LLMs. Based on the framework, we conduct experiments demonstrating two key claims: first, that the computational structure of psychopathology exists in LLMs; and second, that executing this computational structure results in psychopathological functions. We further observe that as LLM size increases, the computational structure of psychopathology becomes denser and that the functions become more effective. Taken together, the empirical results corroborate our hypothesis that network-theoretic computations of psychopathology have already emerged in LLMs. This suggests that certain LLM behaviors mirroring psychopathology may not be a superficial mimicry but a feature of their internal processing. Our work shows the promise of developing a new powerful in silico model of psychopathology and also alludes to the possibility of safety threat from the AI systems with psychopathological behaviors in the near future.
Authors: Lekang Yang, Yuetong Liu, Yitong Zhang, Jia Li
Abstract: Software development relies heavily on extensive unit testing, which makes the efficiency of automated Unit Test Generation (UTG) particularly important. However, most existing LLMs generate test cases one token at a time in each forward pass, which leads to inefficient UTG. Recently, diffusion LLMs (dLLMs) have emerged, offering promising parallel generation capabilities and showing strong potential for efficient UTG. Despite this advantage, their application to UTG is still constrained by a clear trade-off between efficiency and test quality, since increasing the number of tokens generated in each step often causes a sharp decline in the quality of test cases. To overcome this limitation, we present DiffTester, an acceleration framework specifically tailored for dLLMs in UTG. The key idea of DiffTester is that unit tests targeting the same focal method often share repetitive structural patterns. By dynamically identifying these common patterns through abstract syntax tree analysis during generation, DiffTester adaptively increases the number of tokens produced at each step without compromising the quality of the output. To enable comprehensive evaluation, we extend the original TestEval benchmark, which was limited to Python, by introducing additional programming languages including Java and C++. Extensive experiments on three benchmarks with two representative models show that DiffTester delivers significant acceleration while preserving test coverage. Moreover, DiffTester generalizes well across different dLLMs and programming languages, providing a practical and scalable solution for efficient UTG in software development. Code and data are publicly available at https://github.com/wellbeingyang/DLM4UTG-open .
Authors: Shourya Batra, Pierce Tillman, Samarth Gaggar, Shashank Kesineni, Kevin Zhu, Sunishchal Dev, Ashwinee Panda, Vasu Sharma, Maheep Chaudhary
Abstract: As Large Language Models (LLMs) evolve into personal assistants with access to sensitive user data, they face a critical privacy challenge: while prior work has addressed output-level privacy, recent findings reveal that LLMs often leak private information through their internal reasoning processes, violating contextual privacy expectations. These leaky thoughts occur when models inadvertently expose sensitive details in their reasoning traces, even when final outputs appear safe. The challenge lies in preventing such leakage without compromising the model's reasoning capabilities, requiring a delicate balance between privacy and utility. We introduce Steering Activations towards Leakage-free Thinking (SALT), a lightweight test-time intervention that mitigates privacy leakage in model's Chain of Thought (CoT) by injecting targeted steering vectors into hidden state. We identify the high-leakage layers responsible for this behavior. Through experiments across multiple LLMs, we demonstrate that SALT achieves reductions including $18.2\%$ reduction in CPL on QwQ-32B, $17.9\%$ reduction in CPL on Llama-3.1-8B, and $31.2\%$ reduction in CPL on Deepseek in contextual privacy leakage dataset AirGapAgent-R while maintaining comparable task performance and utility. Our work establishes SALT as a practical approach for test-time privacy protection in reasoning-capable language models, offering a path toward safer deployment of LLM-based personal agents.
Authors: Chunqiu Steven Xia, Zhe Wang, Yan Yang, Yuxiang Wei, Lingming Zhang
Abstract: Large Language Models (LLMs) are reshaping almost all industries, including software engineering. In recent years, a number of LLM agents have been proposed to solve real-world software problems. Such software agents are typically equipped with a suite of coding tools and can autonomously decide the next actions to form complete trajectories to solve end-to-end software tasks. While promising, they typically require dedicated design and may still be suboptimal, since it can be extremely challenging and costly to exhaust the entire agent scaffold design space. Recognizing that software agents are inherently software themselves that can be further refined/modified, researchers have proposed a number of self-improving software agents recently, including the Darwin-G\"odel Machine (DGM). Meanwhile, such self-improving agents require costly offline training on specific benchmarks and may not generalize well across different LLMs or benchmarks. In this paper, we propose Live-SWE-agent, the first live software agent that can autonomously and continuously evolve itself on-the-fly during runtime when solving real-world software problems. More specifically, Live-SWE-agent starts with the most basic agent scaffold with only access to bash tools (e.g., mini-SWE-agent), and autonomously evolves its own scaffold implementation while solving real-world software problems. Our evaluation on the widely studied SWE-bench Verified benchmark shows that LIVE-SWE-AGENT can achieve an impressive solve rate of 77.4% without test-time scaling, outperforming all existing software agents, including the best proprietary solution. Moreover, Live-SWE-agent outperforms state-of-the-art manually crafted software agents on the recent SWE-Bench Pro benchmark, achieving the best-known solve rate of 45.8%.