Authors: Pu Zhao, Xuan Shen, Zhenglun Kong, Yixin Shen, Sung-En Chang, Arash Akbari, Timothy Rupprecht, Lei Lu, Enfu Nan, Changdi Yang, Yumei He, Weiyan Shi, Xingchen Xu, Yu Huang, Wei Jiang, Wei Wang, Yue Chen, Yong He, Yanzhi Wang
Abstract: Recently, Large Language Models (LLMs) have undergone a significant transformation, marked by a rapid rise in both their popularity and capabilities. Leading this evolution are proprietary LLMs like GPT-4 and GPT-o1, which have captured widespread attention in the AI community due to their remarkable performance and versatility. Simultaneously, open-source LLMs, such as LLaMA and Mistral, have made great contributions to the ever-increasing popularity of LLMs due to the ease to customize and deploy the models across diverse applications. Moxin 7B is introduced as a fully open-source LLM developed in accordance with the Model Openness Framework, which moves beyond the simple sharing of model weights to embrace complete transparency in training, datasets, and implementation detail, thus fostering a more inclusive and collaborative research environment that can sustain a healthy open-source ecosystem. To further equip Moxin with various capabilities in different tasks, we develop three variants based on Moxin, including Moxin-VLM, Moxin-VLA, and Moxin-Chinese, which target the vision-language, vision-language-action, and Chinese capabilities, respectively. Experiments show that our models achieve superior performance in various evaluations. We adopt open-source framework and open data for the training. We release our models, along with the available data and code to derive these models.
Authors: Sophie Zhao
Abstract: Recent work has shown that transformer-based language models learn rich geometric structure in their embedding spaces, yet the presence of higher-level cognitive organization within these representations remains underexplored. In this work, we investigate whether sentence embeddings encode a graded, hierarchical structure aligned with human-interpretable cognitive or psychological attributes. We construct a dataset of 480 natural-language sentences annotated with continuous ordinal energy scores and discrete tier labels spanning seven ordered cognitive categories. Using fixed sentence embeddings from multiple transformer models, we evaluate the recoverability of these annotations via linear and shallow nonlinear probes. Across models, both continuous scores and tier labels are reliably decodable, with shallow nonlinear probes providing consistent performance gains over linear probes. Lexical TF-IDF baselines perform substantially worse, indicating that the observed structure is not attributable to surface word statistics alone. Nonparametric permutation tests further confirm that probe performance exceeds chance under label-randomization nulls. Qualitative analyses using UMAP visualizations and confusion matrices reveal smooth low-to-high gradients and predominantly adjacent-tier confusions in embedding space. Taken together, these results provide evidence that transformer embedding spaces exhibit a hierarchical geometric organization aligned with human-defined cognitive attributes, while remaining agnostic to claims of internal awareness or phenomenology.
Authors: Shaofei Cai, Yulei Qin, Haojia Lin, Zihan Xu, Gang Li, Yuchen Shi, Zongyi Li, Yong Mao, Siqi Cai, Xiaoyu Tan, Yitao Liang, Ke Li, Xing Sun
Abstract: Agentic reinforcement learning (RL) holds great promise for the development of autonomous agents under complex GUI tasks, but its scalability remains severely hampered by the verification of task completion. Existing task verification is treated as a passive, post-hoc process: a verifier (i.e., rule-based scoring script, reward or critic model, and LLM-as-a-Judge) analyzes the agent's entire interaction trajectory to determine if the agent succeeds. Such processing of verbose context that contains irrelevant, noisy history poses challenges to the verification protocols and therefore leads to prohibitive cost and low reliability. To overcome this bottleneck, we propose SmartSnap, a paradigm shift from this passive, post-hoc verification to proactive, in-situ self-verification by the agent itself. We introduce the Self-Verifying Agent, a new type of agent designed with dual missions: to not only complete a task but also to prove its accomplishment with curated snapshot evidences. Guided by our proposed 3C Principles (Completeness, Conciseness, and Creativity), the agent leverages its accessibility to the online environment to perform self-verification on a minimal, decisive set of snapshots. Such evidences are provided as the sole materials for a general LLM-as-a-Judge verifier to determine their validity and relevance. Experiments on mobile tasks across model families and scales demonstrate that our SmartSnap paradigm allows training LLM-driven agents in a scalable manner, bringing performance gains up to 26.08% and 16.66% respectively to 8B and 30B models. The synergizing between solution finding and evidence seeking facilitates the cultivation of efficient, self-verifying agents with competitive performance against DeepSeek V3.1 and Qwen3-235B-A22B.
Authors: Zubaida Mohammed Albadani, Mohammed Q. Shormani
Abstract: This study investigates the syntax of qulk-clauses in Yemeni Ibbi Arabic (YIA) within the Minimalist Program. The construction qulk-clause, a morphologically fused form meaning 'I said,' introduces embedded declarative interrogative, and imperative clauses, often eithout complementizer. The central proposal of this paper is that qulk-clauses are biclausal structures in which qulk functions a clause-embedding predicate sec;ecting a dull CP complement. By applying core minimalist operations, viz., Merge, Move, Agree, and Spell-out, the study provides a layered syntactic analysis of qulk-clauses, for illustrating how their derivation proceeds through standard computational steps and post-syntactic processes such as Morphological Merger. The proposal also accounts for dialect-specific features like bipartite negation, cliticization, and CP embedding. The findings offer theoretical contributions to generative syntax, specifically minimalism. The study concludes raising theoretical questions concerning extending the analysis to the addressee-clause kil-k 'you said'. It also provides insights into the possibility of the universality of minimalism.
Authors: Donggyun Bae, Jongil Park
Abstract: We propose a novel framework, termed Fourier-Activated Adapter (FAA), for parameter-efficient fine-tuning of large pre-trained language models. By incorporating random Fourier features into lightweight adapter modules, FAA decomposes intermediate representations into complementary low- and high-frequency components, enabling frequency-aware modulation of semantic information. This design allows the model to selectively emphasize informative frequency bands during adaptation while preserving the representational capacity of the frozen backbone. Extensive experiments on GLUE, E2E NLG, and instruction-tuning benchmarks demonstrate that FAA consistently achieves competitive or superior performance compared to existing parameter-efficient fine-tuning methods, while maintaining low computational and memory overhead. Ablation studies further verify the effectiveness of frequency-aware activation and adaptive weighting mechanisms, highlighting FAA as a robust and efficient approach for post-training large language models.
Authors: Elsen Ronando, Sozo Inoue
Abstract: In this paper, we propose an LLM-Guided Exemplar Selection framework to address a key limitation in state-of-the-art Human Activity Recognition (HAR) methods: their reliance on large labeled datasets and purely geometric exemplar selection, which often fail to distinguish similar weara-ble sensor activities such as walking, walking upstairs, and walking downstairs. Our method incorporates semantic reasoning via an LLM-generated knowledge prior that captures feature importance, inter-class confusability, and exemplar budget multipliers, and uses it to guide exemplar scoring and selection. These priors are combined with margin-based validation cues, PageRank centrality, hubness penalization, and facility-location optimization to obtain a compact and informative set of exemplars. Evaluated on the UCI-HAR dataset under strict few-shot conditions, the framework achieves a macro F1-score of 88.78%, outperforming classical approaches such as random sampling, herding, and $k$-center. The results show that LLM-derived semantic priors, when integrated with structural and geometric cues, provide a stronger foundation for selecting representative sensor exemplars in few-shot wearable-sensor HAR.
Authors: Chenggong Zhang, Haopeng Wang
Abstract: Hallucinations in Large Language Models (LLMs) pose a significant challenge, generating misleading or unverifiable content that undermines trust and reliability. Existing evaluation methods, such as KnowHalu, employ multi-stage verification but suffer from high computational costs. To address this, we integrate the Hughes Hallucination Evaluation Model (HHEM), a lightweight classification-based framework that operates independently of LLM-based judgments, significantly improving efficiency while maintaining high detection accuracy. We conduct a comparative analysis of hallucination detection methods across various LLMs, evaluating True Positive Rate (TPR), True Negative Rate (TNR), and Accuracy on question-answering (QA) and summarization tasks. Our results show that HHEM reduces evaluation time from 8 hours to 10 minutes, while HHEM with non-fabrication checking achieves the highest accuracy \(82.2\%\) and TPR \(78.9\%\). However, HHEM struggles with localized hallucinations in summarization tasks. To address this, we introduce segment-based retrieval, improving detection by verifying smaller text components. Additionally, our cumulative distribution function (CDF) analysis indicates that larger models (7B-9B parameters) generally exhibit fewer hallucinations, while intermediate-sized models show higher instability. These findings highlight the need for structured evaluation frameworks that balance computational efficiency with robust factual validation, enhancing the reliability of LLM-generated content.
Authors: Cattalyya Nuengsigkapian
Abstract: Retrieval-Augmented Generation (RAG) in open-domain settings faces significant challenges regarding irrelevant information in retrieved documents and the alignment of generated answers with user intent. We present HiFi-RAG (Hierarchical Filtering RAG), the winning closed-source system in the Text-to-Text static evaluation of the MMU-RAGent NeurIPS 2025 Competition. Our approach moves beyond standard embedding-based retrieval via a multi-stage pipeline. We leverage the speed and cost-efficiency of Gemini 2.5 Flash (4-6x cheaper than Pro) for query formulation, hierarchical content filtering, and citation attribution, while reserving the reasoning capabilities of Gemini 2.5 Pro for final answer generation. On the MMU-RAGent validation set, our system outperformed the baseline, improving ROUGE-L to 0.274 (+19.6%) and DeBERTaScore to 0.677 (+6.2%). On Test2025, our custom dataset evaluating questions that require post-cutoff knowledge (post January 2025), HiFi-RAG outperforms the parametric baseline by 57.4% in ROUGE-L and 14.9% in DeBERTaScore.
Authors: Jie Zhou, Xin Chen, Jie Zhang, Zhe Li
Abstract: Large Language Models (LLMs) are reshaping learning paradigms, cognitive processes, and research methodologies across a wide range of domains. Integrating LLMs with professional fields and redefining the relationship between LLMs and domain-specific applications has become a critical challenge for promoting enterprise digital transformation and broader social development. To effectively integrate LLMs into the accounting domain, it is essential to understand their domain-specific reasoning capabilities. This study introduces the concept of vertical-domain accounting reasoning and establishes evaluation criteria by analyzing the training data characteristics of representative GLM-series models. These criteria provide a foundation for subsequent research on reasoning paradigms and offer benchmarks for improving accounting reasoning performance. Based on this framework, we evaluate several representative models, including GLM-6B, GLM-130B, GLM-4, and OpenAI GPT-4, on a set of accounting reasoning tasks. Experimental results show that different prompt engineering strategies lead to varying degrees of performance improvement across models, with GPT-4 achieving the strongest accounting reasoning capability. However, current LLMs still fall short of real-world application requirements. In particular, further optimization is needed for deployment in enterprise-level accounting scenarios to fully realize the potential value of LLMs in this domain.
Authors: Jungyeul Park, Chulwoo Park
Abstract: The design of Korean constituency treebanks raises a fundamental representational question concerning the choice of terminal units. Although Korean words are morphologically complex, treating morphemes as constituency terminals conflates word internal morphology with phrase level syntactic structure and creates mismatches with eojeol based dependency resources. This paper argues for an eojeol based constituency representation, with morphological segmentation and fine grained part of speech information encoded in a separate, non constituent layer. A comparative analysis shows that, under explicit normalization assumptions, the Sejong and Penn Korean treebanks can be treated as representationally equivalent at the eojeol based constituency level. Building on this result, we outline an eojeol based annotation scheme that preserves interpretable constituency and supports cross treebank comparison and constituency dependency conversion.
Authors: Suhua Wang, Zifan Wang, Xiaoxin Sun, D. J. Wang, Zhanbo Liu, Xin Li
Abstract: As an endangered language, Manchu presents unique challenges for speech synthesis, including severe data scarcity and strong phonological agglutination. This paper proposes ManchuTTS(Manchu Text to Speech), a novel approach tailored to Manchu's linguistic characteristics. To handle agglutination, this method designs a three-tier text representation (phoneme, syllable, prosodic) and a cross-modal hierarchical attention mechanism for multi-granular alignment. The synthesis model integrates deep convolutional networks with a flow-matching Transformer, enabling efficient, non-autoregressive generation. This method further introduce a hierarchical contrastive loss to guide structured acoustic-linguistic correspondence. To address low-resource constraints, This method construct the first Manchu TTS dataset and employ a data augmentation strategy. Experiments demonstrate that ManchuTTS attains a MOS of 4.52 using a 5.2-hour training subset derived from our full 6.24-hour annotated corpus, outperforming all baseline models by a notable margin. Ablations confirm hierarchical guidance improves agglutinative word pronunciation accuracy (AWPA) by 31% and prosodic naturalness by 27%.
Authors: Xuan Luo, Kailai Zhang, Xifeng Yan
Abstract: When reading books, humans focus primarily on the current page, flipping back to recap prior context only when necessary. Similarly, we demonstrate that Large Language Models (LLMs) can learn to dynamically determine when to attend to global context. We propose All-or-Here Attention (AHA), which utilizes a binary router per attention head to dynamically toggle between full attention and local sliding window attention for each token. Our results indicate that with a window size of 256 tokens, up to 93\% of the original full attention operations can be replaced by sliding window attention without performance loss. Furthermore, by evaluating AHA across various window sizes, we identify a long-tail distribution in context dependency, where the necessity for full attention decays rapidly as the local window expands. By decoupling local processing from global access, AHA reveals that full attention is largely redundant, and that efficient inference requires only on-demand access to the global context.
Authors: Zhiqiang Gao, Shihao Gao, Zixing Zhang, Yihao Guo, Hongyu Chen, Jing Han
Abstract: Understanding sentiment in multimodal conversations is a complex yet crucial challenge toward building emotionally intelligent AI systems. The Multimodal Conversational Aspect-based Sentiment Analysis (MCABSA) Challenge invited participants to tackle two demanding subtasks: (1) extracting a comprehensive sentiment sextuple, including holder, target, aspect, opinion, sentiment, and rationale from multi-speaker dialogues, and (2) detecting sentiment flipping, which detects dynamic sentiment shifts and their underlying triggers. For Subtask-I, in the present paper, we designed a structured prompting pipeline that guided large language models (LLMs) to sequentially extract sentiment components with refined contextual understanding. For Subtask-II, we further leveraged the complementary strengths of three LLMs through ensembling to robustly identify sentiment transitions and their triggers. Our system achieved a 47.38% average score on Subtask-I and a 74.12% exact match F1 on Subtask-II, showing the effectiveness of step-wise refinement and ensemble strategies in rich, multimodal sentiment analysis tasks.
Authors: Chen Su, Yuanhe Tian, Yan Song
Abstract: With the advancement of large language models (LLMs), diverse time series analysis tasks are reformulated as time series question answering (TSQA) through a unified natural language interface. However, existing LLM-based approaches largely adopt general natural language processing techniques and are prone to reasoning errors when handling complex numerical sequences. Different from purely textual tasks, time series data are inherently verifiable, enabling consistency checking between reasoning steps and the original input. Motivated by this property, we propose T3LLM, which performs multi-step reasoning with an explicit correction mechanism for time series question answering. The T3LLM framework consists of three LLMs, namely, a worker, a reviewer, and a student, that are responsible for generation, review, and reasoning learning, respectively. Within this framework, the worker generates step-wise chains of thought (CoT) under structured prompts, while the reviewer inspects the reasoning, identifies erroneous steps, and provides corrective comments. The collaboratively generated corrected CoT are used to fine-tune the student model, internalizing multi-step reasoning and self-correction into its parameters. Experiments on multiple real-world TSQA benchmarks demonstrate that T3LLM achieves state-of-the-art performance over strong LLM-based baselines.
Authors: Fanglin Xu, Wei Zhang, Jian Yang, Guo Chen, Aishan Liu, Zhoujun Li, Xianglong Liu, Bryan Dai
Abstract: The rapid advancement of code large language models (LLMs) has sparked significant research interest in systematically evaluating their code generation capabilities, yet existing benchmarks predominantly assess models at a single structural granularity and focus on limited programming languages, obscuring fine-grained capability variations across different code scopes and multilingual scenarios. We introduce M2G-Eval, a multi-granularity, multilingual framework for evaluating code generation in large language models (LLMs) across four levels: Class, Function, Block, and Line. Spanning 18 programming languages, M2G-Eval includes 17K+ training tasks and 1,286 human-annotated, contamination-controlled test instances. We develop M2G-Eval-Coder models by training Qwen3-8B with supervised fine-tuning and Group Relative Policy Optimization. Evaluating 30 models (28 state-of-the-art LLMs plus our two M2G-Eval-Coder variants) reveals three main findings: (1) an apparent difficulty hierarchy, with Line-level tasks easiest and Class-level most challenging; (2) widening performance gaps between full- and partial-granularity languages as task complexity increases; and (3) strong cross-language correlations, suggesting that models learn transferable programming concepts. M2G-Eval enables fine-grained diagnosis of code generation capabilities and highlights persistent challenges in synthesizing complex, long-form code.
Authors: Ziqi Jin, Bin Wang, Xiang Lin, Lidong Bing, Aixin Sun
Abstract: Diffusion models offer appealing properties for language generation, such as parallel decoding and iterative refinement, but the discrete and highly structured nature of text challenges the direct application of diffusion principles. In this paper, we revisit diffusion language modeling from the view of diffusion process and language modeling, and outline five properties that separate diffusion mechanics from language-specific requirements. We first categorize existing approaches into continuous diffusion in embedding space and discrete diffusion over tokens. We then show that each satisfies only part of the five essential properties and therefore reflects a structural trade-off. Through analyses of recent large diffusion language models, we identify two central issues: (i) uniform corruption does not respect how information is distributed across positions, and (ii) token-wise marginal training cannot capture multi-token dependencies during parallel decoding. These observations motivate diffusion processes that align more closely with the structure of text, and encourage future work toward more coherent diffusion language models.
Authors: Hadi Mohammadi, Tamas Kozak, Anastasia Giachanou
Abstract: Chain-of-thought (CoT) reasoning has emerged as a powerful technique for improving the problem-solving capabilities of large language models (LLMs), particularly for tasks requiring multi-step reasoning. However, recent studies show that CoT explanations often fail to reflect the model's actual reasoning process, as models may produce coherent yet misleading justifications or modify answers without acknowledging external cues. Such discrepancies undermine the reliability of CoT-based methods for safety supervision and alignment monitoring, as models can generate plausible but deceptive rationales for incorrect answers. To better understand this limitation, we evaluate two optimization methods, Group Relative Policy Optimization (GRPO) and Direct Preference Optimization (DPO), in their ability to improve CoT faithfulness. Our experiments show that GRPO achieves higher performance than DPO in larger models, with the Qwen2.5-14B-Instruct model attaining the best results across all evaluation metrics. Both approaches exhibit positive correlations between model size and performance, but GRPO shows greater potential for improving faithfulness metrics, albeit with less stable behavior at smaller scales. These results suggest that GRPO offers a promising direction for developing more transparent and trustworthy reasoning in LLMs.
Authors: Pere Martra
Abstract: Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities. While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models), and multi-step reasoning remains robust (MUSR). This pattern challenges the prevailing assumption that pruning induces uniform degradation. We evaluated seven expansion ratio configurations using comprehensive benchmarks assessing factual knowledge, mathematical reasoning, language comprehension, instruction-following, and truthfulness. Our analysis identifies the expansion ratio as a critical architectural parameter that selectively modulates cognitive capabilities, rather than merely serving as a compression metric. We provide the first systematic characterization of this selective preservation phenomenon. Notably, we document a robust inverse correlation (r = -0.864, p = 0.012 in Llama-3B) between factual knowledge capacity (MMLU) and truthfulness metrics (TruthfulQA-MC2): as knowledge degrades, the model's ability to discriminate misconceptions improves consistently. This connects two previously distinct research areas, demonstrating that MAW-guided width pruning acts as a selective filter, reducing parametric knowledge while preserving or enhancing behavioral alignment. Additionally, we quantify context-dependent efficiency trade-offs: pruned configurations achieve up to 23% reduction in energy consumption (J/token) but incur penalties in single-request latency, whereas batch processing workloads benefit uniformly.
Authors: Yoshith Roy Kotla, Varshith Roy Kotla
Abstract: Deploying large language models (LLMs) in high-stakes domains requires rigorous uncertainty quantification, yet standard softmax probabilities are often poorly calibrated. We present a systematic study of Adaptive Prediction Sets (APS) applied to next-token prediction in transformer-based models with large vocabularies (greater than 250,000 tokens). Our central contribution is the identification of a coverage-efficiency tradeoff: while naive conformal prediction achieves valid coverage, it produces prediction sets of hundreds of tokens, rendering them uninformative. We propose Vocabulary-Aware Conformal Prediction (VACP), a framework that leverages semantic masking and temperature-adjusted scoring to reduce the effective prediction space while provably maintaining marginal coverage. Experiments on Gemma-2B using SQUAD and WikiText benchmarks demonstrate that VACP achieves 89.7 percent empirical coverage (90 percent target) while reducing the mean prediction set size from 847 tokens to 4.3 tokens -- a 197x improvement in efficiency. We provide a theoretical analysis of vocabulary reduction and release our implementation for reproducibility.
Authors: Ahmed Abdullah, Sana Fatima, Haroon Mahmood
Abstract: Hope speech has been relatively underrepresented in Natural Language Processing (NLP). Current studies are largely focused on English, which has resulted in a lack of resources for low-resource languages such as Urdu. As a result, the creation of tools that facilitate positive online communication remains limited. Although transformer-based architectures have proven to be effective in detecting hate and offensive speech, little has been done to apply them to hope speech or, more generally, to test them across a variety of linguistic settings. This paper presents a multilingual framework for hope speech detection with a focus on Urdu. Using pretrained transformer models such as XLM-RoBERTa, mBERT, EuroBERT, and UrduBERT, we apply simple preprocessing and train classifiers for improved results. Evaluations on the PolyHope-M 2025 benchmark demonstrate strong performance, achieving F1-scores of 95.2% for Urdu binary classification and 65.2% for Urdu multi-class classification, with similarly competitive results in Spanish, German, and English. These results highlight the possibility of implementing existing multilingual models in low-resource environments, thus making it easier to identify hope speech and helping to build a more constructive digital discourse.
Authors: Anaelia Ovalle, Candace Ross, Sebastian Ruder, Adina Williams, Karen Ullrich, Mark Ibrahim, Levent Sagun
Abstract: Large language models demonstrate strong reasoning capabilities through chain-of-thought prompting, but whether this reasoning quality transfers across languages remains underexplored. We introduce a human-validated framework to evaluate whether model-generated reasoning traces logically support their conclusions across languages. Analyzing 65k reasoning traces from GlobalMMLU questions across 6 languages and 6 frontier models, we uncover a critical blind spot: while models achieve high task accuracy, their reasoning can fail to support their conclusions. Reasoning traces in non-Latin scripts show at least twice as much misalignment between their reasoning and conclusions than those in Latin scripts. We develop an error taxonomy through human annotation to characterize these failures, finding they stem primarily from evidential errors (unsupported claims, ambiguous facts) followed by illogical reasoning steps. Our findings demonstrate that current multilingual evaluation practices provide an incomplete picture of model reasoning capabilities and highlight the need for reasoning-aware evaluation frameworks.
Authors: Sashank Chapala, Maksym Mironov, Songgaojun Deng
Abstract: Large Language Models (LLMs) are increasingly used to simulate population responses, a method known as ``Silicon Sampling''. However, responses to socially sensitive questions frequently exhibit Social Desirability Bias (SDB), diverging from real human data toward socially acceptable answers. Existing studies on social desirability bias in LLM-based sampling remain limited. In this work, we investigate whether minimal, psychologically grounded prompt wording can mitigate this bias and improve alignment between silicon and human samples. We conducted a study using data from the American National Election Study (ANES) on three LLMs from two model families: the open-source Llama-3.1 series and GPT-4.1-mini. We first replicate a baseline silicon sampling study, confirming the persistent Social Desirability Bias. We then test four prompt-based mitigation methods: \emph{reformulated} (neutral, third-person phrasing), \emph{reverse-coded} (semantic inversion), and two meta-instructions, \emph{priming} and \emph{preamble}, respectively encouraging analytics and sincerity. Alignment with ANES is evaluated using Jensen-Shannon Divergence with bootstrap confidence intervals. Our results demonstrate that reformulated prompts most effectively improve alignment by reducing distribution concentration on socially acceptable answers and achieving distributions closer to ANES. Reverse-coding produced mixed results across eligible items, while the Priming and Preamble encouraged response uniformity and showed no systematic benefit for bias mitigation. Our findings validate the efficacy of prompt-based framing controls in mitigating inherent Social Desirability Bias in LLMs, providing a practical path toward more representative silicon samples.
Authors: Md Badsha Biswas
Abstract: Infant mortality remains a significant public health concern in the United States, with birth defects identified as a leading cause. Despite ongoing efforts to understand the causes of negative pregnancy outcomes like miscarriage, stillbirths, birth defects, and premature birth, there is still a need for more comprehensive research and strategies for intervention. This paper introduces a novel approach that uses publicly available social media data, especially from platforms like Twitter, to enhance current datasets for studying negative pregnancy outcomes through observational research. The inherent challenges in utilizing social media data, including imbalance, noise, and lack of structure, necessitate robust preprocessing techniques and data augmentation strategies. By constructing a natural language processing (NLP) pipeline, we aim to automatically identify women sharing their pregnancy experiences, categorizing them based on reported outcomes. Women reporting full gestation and normal birth weight will be classified as positive cases, while those reporting negative pregnancy outcomes will be identified as negative cases. Furthermore, this study offers potential applications in assessing the causal impact of specific interventions, treatments, or prenatal exposures on maternal and fetal health outcomes. Additionally, it provides a framework for future health studies involving pregnant cohorts and comparator groups. In a broader context, our research showcases the viability of social media data as an adjunctive resource in epidemiological investigations about pregnancy outcomes.
Authors: Aiwei Liu, Minghua He, Shaoxun Zeng, Sijun Zhang, Linhao Zhang, Chuhan Wu, Wei Jia, Yuan Liu, Xiao Zhou, Jie Zhou
Abstract: Autoregressive (AR) generation is the standard decoding paradigm for Large Language Models (LLMs), but its token-by-token nature limits parallelism at inference time. Diffusion Language Models (DLLMs) offer parallel decoding by recovering multiple masked tokens per step; however, in practice they often fail to translate this parallelism into deployment speed gains over optimized AR engines (e.g., vLLM). A key reason is that many DLLMs rely on bidirectional attention, which breaks standard prefix KV caching and forces repeated contextualization, undermining efficiency. We propose WeDLM, a diffusion decoding framework built entirely on standard causal attention to make parallel generation prefix-cache friendly. The core idea is to let each masked position condition on all currently observed tokens while keeping a strict causal mask, achieved by Topological Reordering that moves observed tokens to the physical prefix while preserving their logical positions. Building on this property, we introduce a streaming decoding procedure that continuously commits confident tokens into a growing left-to-right prefix and maintains a fixed parallel workload, avoiding the stop-and-wait behavior common in block diffusion methods. Experiments show that WeDLM preserves the quality of strong AR backbones while delivering substantial speedups, approaching 3x on challenging reasoning benchmarks and up to 10x in low-entropy generation regimes; critically, our comparisons are against AR baselines served by vLLM under matched deployment settings, demonstrating that diffusion-style decoding can outperform an optimized AR engine in practice.
Authors: Jian Chen, Leilei Su, Cong Sun
Abstract: Background and Objective: Biomedical Named Entity Recognition (BioNER) is a foundational task in medical informatics, crucial for downstream applications like drug discovery and clinical trial matching. However, adapting general-domain Large Language Models (LLMs) to this task is often hampered by their lack of domain-specific knowledge and the performance degradation caused by low-quality training data. To address these challenges, we introduce BioSelectTune, a highly efficient, data-centric framework for fine-tuning LLMs that prioritizes data quality over quantity. Methods and Results: BioSelectTune reformulates BioNER as a structured JSON generation task and leverages our novel Hybrid Superfiltering strategy, a weak-to-strong data curation method that uses a homologous weak model to distill a compact, high-impact training dataset. Conclusions: Through extensive experiments, we demonstrate that BioSelectTune achieves state-of-the-art (SOTA) performance across multiple BioNER benchmarks. Notably, our model, trained on only 50% of the curated positive data, not only surpasses the fully-trained baseline but also outperforms powerful domain-specialized models like BioMedBERT.
Authors: Dongning Rao, Yunbiao Zeng, Zhihua Jiang, Jujian Lv
Abstract: Human-interaction-involved applications underscore the need for Multi-modal Sentiment Analysis (MSA). Although many approaches have been proposed to address the subtle emotions in different modalities, the power of explanations and temporal alignments is still underexplored. Thus, this paper proposes the Text-routed sparse mixture-of-Experts model with eXplanation and Temporal alignment for MSA (TEXT). TEXT first augments explanations for MSA via Multi-modal Large Language Models (MLLM), and then novelly aligns the epresentations of audio and video through a temporality-oriented neural network block. TEXT aligns different modalities with explanations and facilitates a new text-routed sparse mixture-of-experts with gate fusion. Our temporal alignment block merges the benefits of Mamba and temporal cross-attention. As a result, TEXT achieves the best performance cross four datasets among all tested models, including three recently proposed approaches and three MLLMs. TEXT wins on at least four metrics out of all six metrics. For example, TEXT decreases the mean absolute error to 0.353 on the CH-SIMS dataset, which signifies a 13.5% decrement compared with recently proposed approaches.
Authors: Muhammad Zain Ali, Bernhard Pfahringer, Tony Smith
Abstract: Misinformation on social media is a widely acknowledged issue, and researchers worldwide are actively engaged in its detection. However, low-resource languages such as Urdu have received limited attention in this domain. An obvious approach is to utilize a multilingual pretrained language model and fine-tune it for a downstream classification task, such as misinformation detection. However, these models struggle with domain-specific terms, leading to suboptimal performance. To address this, we investigate the effectiveness of domain adaptation before fine-tuning for fake news classification in Urdu, employing a staged training approach to optimize model generalization. We evaluate two widely used multilingual models, XLM-RoBERTa and mBERT, and apply domain-adaptive pretraining using a publicly available Urdu news corpus. Experiments on four publicly available Urdu fake news datasets show that domain-adapted XLM-R consistently outperforms its vanilla counterpart, while domain-adapted mBERT exhibits mixed results.
Authors: Risha Surana, Adrian Law, Sunwoo Kim, Rishab Sridhar, Angxiao Han, Peiyu Hong
Abstract: Clinical notes are often stored in unstructured or semi-structured formats after extraction from electronic medical record (EMR) systems, which complicates their use for secondary analysis and downstream clinical applications. Reliable identification of section boundaries is a key step toward structuring these notes, as sections such as history of present illness, medications, and discharge instructions each provide distinct clinical contexts. In this work, we evaluate rule-based baselines, domain-specific transformer models, and large language models for clinical note segmentation using a curated dataset of 1,000 notes from MIMIC-IV. Our experiments show that large API-based models achieve the best overall performance, with GPT-5-mini reaching a best average F1 of 72.4 across sentence-level and freetext segmentation. Lightweight baselines remain competitive on structured sentence-level tasks but falter on unstructured freetext. Our results provide guidance for method selection and lay the groundwork for downstream tasks such as information extraction, cohort identification, and automated summarization.
Authors: Sameer Sitoula, Tej Bahadur Shahi, Laxmi Prasad Bhatt, Anisha Pokhrel, Arjun Neupane
Abstract: Social media (SM) platforms (e.g. Facebook, Twitter, and Reddit) are increasingly leveraged to share opinions and emotions, specifically during challenging events, such as natural disasters, pandemics, and political elections, and joyful occasions like festivals and celebrations. Among the SM platforms, Reddit provides a unique space for its users to anonymously express their experiences and thoughts on sensitive issues such as health and daily life. In this work, we present a novel dataset, called NepEMO, for multi-label emotion (MLE) and sentiment classification (SC) on the Nepali subreddit post. We curate and build a manually annotated dataset of 4,462 posts (January 2019- June 2025) written in English, Romanised Nepali and Devanagari script for five emotions (fear, anger, sadness, joy, and depression) and three sentiment classes (positive, negative, and neutral). We perform a detailed analysis of posts to capture linguistic insights, including emotion trends, co-occurrence of emotions, sentiment-specific n-grams, and topic modelling using Latent Dirichlet Allocation and TF-IDF keyword extraction. Finally, we compare various traditional machine learning (ML), deep learning (DL), and transformer models for MLE and SC tasks. The result shows that transformer models consistently outperform the ML and DL models for both tasks.
Authors: Shihao Cai, Runnan Fang, Jialong Wu, Baixuan Li, Xinyu Wang, Yong Jiang, Liangcai Su, Liwen Zhang, Wenbiao Yin, Zhen Zhang, Fuli Feng, Pengjun Xie, Xiaobin Wang
Abstract: Conducting reinforcement learning (RL) in simulated environments offers a cost-effective and highly scalable way to enhance language-based agents. However, previous work has been limited to semi-automated environment synthesis or tasks lacking sufficient difficulty, offering little breadth or depth. In addition, the instability of simulated users integrated into these environments, along with the heterogeneity across simulated environments, poses further challenges for agentic RL. In this work, we propose: (1) a unified pipeline for automated and scalable synthesis of simulated environments associated with high-difficulty but easily verifiable tasks; and (2) an environment level RL algorithm that not only effectively mitigates user instability but also performs advantage estimation at the environment level, thereby improving training efficiency and stability. Comprehensive evaluations on agentic benchmarks, including tau-bench, tau2-Bench, and VitaBench, validate the effectiveness of our proposed method. Further in-depth analyses underscore its out-of-domain generalization.
Authors: Haoyuan Wu, Hai Wang, Jiajia Wu, Jinxiang Ou, Keyao Wang, Weile Chen, Zihao Zheng, Bei Yu
Abstract: Recent advancements have shown that reinforcement learning (RL) can substantially improve the reasoning abilities of large language models (LLMs). The effectiveness of such RL training, however, depends critically on the exploration space defined by the pre-trained model's token-output distribution. In this paper, we revisit the standard cross-entropy loss, interpreting it as a specific instance of policy gradient optimization applied within a single-step episode. To systematically study how the pre-trained distribution shapes the exploration potential for subsequent RL, we propose a generalized pre-training objective that adapts on-policy RL principles to supervised learning. By framing next-token prediction as a stochastic decision process, we introduce a reward-shaping strategy that explicitly balances diversity and precision. Our method employs a positive reward scaling factor to control probability concentration on ground-truth tokens and a rank-aware mechanism that treats high-ranking and low-ranking negative tokens asymmetrically. This allows us to reshape the pre-trained token-output distribution and investigate how to provide a more favorable exploration space for RL, ultimately enhancing end-to-end reasoning performance. Contrary to the intuition that higher distribution entropy facilitates effective exploration, we find that imposing a precision-oriented prior yields a superior exploration space for RL.
Authors: Mengdi Chai, Ali R. Zomorrodi
Abstract: Large Language Models (LLMs) have demonstrated promise in medical knowledge assessments, yet their practical utility in real-world clinical decision-making remains underexplored. In this study, we evaluated the performance of three state-of-the-art LLMs-ChatGPT-4o, Gemini 1.5 Pro, and LIama 3.3 70B-in clinical decision support across the entire clinical reasoning workflow of a typical patient encounter. Using 36 case studies, we first assessed LLM's out-of-the-box performance across five key sequential clinical decision-making tasks under two temperature settings (default vs. zero): differential diagnosis, essential immediate steps, relevant diagnostic testing, final diagnosis, and treatment recommendation. All models showed high variability by task, achieving near-perfect accuracy in final diagnosis, poor performance in relevant diagnostic testing, and moderate performance in remaining tasks. Furthermore, ChatGPT performed better under the zero temperature, whereas LIama showed stronger performance under the default temperature. Next, we assessed whether prompt engineering could enhance LLM performance by applying variations of the MedPrompt framework, incorporating targeted and random dynamic few-shot learning. The results demonstrate that prompt engineering is not a one-size-fit-all solution. While it significantly improved the performance on the task with lowest baseline accuracy (relevant diagnostic testing), it was counterproductive for others. Another key finding was that the targeted dynamic few-shot prompting did not consistently outperform random selection, indicating that the presumed benefits of closely matched examples may be counterbalanced by loss of broader contextual diversity. These findings suggest that the impact of prompt engineering is highly model and task-dependent, highlighting the need for tailored, context-aware strategies for integrating LLMs into healthcare.
Authors: Tao Yu, Yongqi An, Kuan Zhu, Guibo Zhu, Ming Tang, Jinqiao Wang
Abstract: Large Language Models (LLMs) demonstrate impressive performance across natural language tasks but incur substantial computational and storage costs due to their scale. Post-training structured pruning offers an efficient solution. However, when few-shot calibration sets fail to adequately reflect the pretraining data distribution, existing methods exhibit limited generalization to downstream tasks. To address this issue, we propose Function-Aware Neuron Grouping (FANG), a post-training pruning framework that alleviates calibration bias by identifying and preserving neurons critical to specific function. FANG groups neurons with similar function based on the type of semantic context they process and prunes each group independently. During importance estimation within each group, tokens that strongly correlate with the functional role of the neuron group are given higher weighting. Additionally, FANG also preserves neurons that contribute across multiple context types. To achieve a better trade-off between sparsity and performance, it allocates sparsity to each block adaptively based on its functional complexity. Experiments show that FANG improves downstream accuracy while preserving language modeling performance. It achieves the state-of-the-art (SOTA) results when combined with FLAP and OBC, two representative pruning methods. Specifically, FANG outperforms FLAP and OBC by 1.5%--8.5% in average accuracy under 30% and 40% sparsity.
Authors: Wenxuan Xu, Arvind Pillai, Subigya Nepal, Amanda C Collins, Daniel M Mackin, Michael V Heinz, Tess Z Griffin, Nicholas C Jacobson, Andrew Campbell
Abstract: Multimodal health sensing offers rich behavioral signals for assessing mental health, yet translating these numerical time-series measurements into natural language remains challenging. Current LLMs cannot natively ingest long-duration sensor streams, and paired sensor-text datasets are scarce. To address these challenges, we introduce LENS, a framework that aligns multimodal sensing data with language models to generate clinically grounded mental-health narratives. LENS first constructs a large-scale dataset by transforming Ecological Momentary Assessment (EMA) responses related to depression and anxiety symptoms into natural-language descriptions, yielding over 100,000 sensor-text QA pairs from 258 participants. To enable native time-series integration, we train a patch-level encoder that projects raw sensor signals directly into an LLM's representation space. Our results show that LENS outperforms strong baselines on standard NLP metrics and task-specific measures of symptom-severity accuracy. A user study with 13 mental-health professionals further indicates that LENS-produced narratives are comprehensive and clinically meaningful. Ultimately, our approach advances LLMs as interfaces for health sensing, providing a scalable path toward models that can reason over raw behavioral signals and support downstream clinical decision-making.
Authors: Kerem Zaman, Shashank Srivastava
Abstract: Recent work, using the Biasing Features metric, labels a CoT as unfaithful if it omits a prompt-injected hint that affected the prediction. We argue this metric confuses unfaithfulness with incompleteness, the lossy compression needed to turn distributed transformer computation into a linear natural language narrative. On multi-hop reasoning tasks with Llama-3 and Gemma-3, many CoTs flagged as unfaithful by Biasing Features are judged faithful by other metrics, exceeding 50% in some models. With a new faithful@k metric, we show that larger inference-time token budgets greatly increase hint verbalization (up to 90% in some settings), suggesting much apparent unfaithfulness is due to tight token limits. Using Causal Mediation Analysis, we further show that even non-verbalized hints can causally mediate prediction changes through the CoT. We therefore caution against relying solely on hint-based evaluations and advocate a broader interpretability toolkit, including causal mediation and corruption-based metrics.
Authors: TJ Bai, Jason Eisner
Abstract: Large language models are increasingly deployed in multi-agent workflows. We introduce Prompt Choreography, a framework that efficiently executes LLM workflows by maintaining a dynamic, global KV cache. Each LLM call can attend to an arbitrary, reordered subset of previously encoded messages. Parallel calls are supported. Though caching messages' encodings sometimes gives different results from re-encoding them in a new context, we show in diverse settings that fine-tuning the LLM to work with the cache can help it mimic the original results. Prompt Choreography significantly reduces per-message latency (2.0--6.2$\times$ faster time-to-first-token) and achieves substantial end-to-end speedups ($>$2.2$\times$) in some workflows dominated by redundant computation.
Authors: Melik\c{s}ah T\"urker, A. Ebrar K{\i}z{\i}lo\u{g}lu, Onur G\"ung\"or, Susan \"Usk\"udarl{\i}
Abstract: Since the inception of BERT, encoder-only Transformers have evolved significantly in computational efficiency, training stability, and long-context modeling. ModernBERT consolidates these advances by integrating Rotary Positional Embeddings (RoPE), FlashAttention, and refined normalization. Despite these developments, Turkish NLP lacks a monolingual encoder trained from scratch incorporating such modern architectural paradigms. This work introduces TabiBERT, a monolingual Turkish encoder based on ModernBERT architecture trained from scratch on a large, curated corpus. TabiBERT is pre-trained on one trillion tokens sampled from an 84.88B token multi-domain corpus: web text (73%), scientific publications (20%), source code (6%), and mathematical content (0.3%). The model supports 8,192-token context length (16x original BERT), achieves up to 2.65x inference speedup, and reduces GPU memory consumption, enabling larger batch sizes. We introduce TabiBench with 28 datasets across eight task categories with standardized splits and protocols, evaluated using GLUE-style macro-averaging. TabiBERT attains 77.58 on TabiBench, outperforming BERTurk by 1.62 points and establishing state-of-the-art on five of eight categories: question answering (+9.55), code retrieval (+2.41), and document retrieval (+0.60). Compared with task-specific prior best results, including specialized models like TurkishBERTweet, TabiBERT achieves +1.47 average improvement, indicating robust cross-domain generalization. We release model weights, training configurations, and evaluation code for transparent, reproducible Turkish encoder research.
Authors: Takumi Shiratsuchi, Yuichiro Tanaka, Hakaru Tamukoh
Abstract: Large language models (LLMs) have achieved state-of-the-art performance in natural language processing; however, their high computational cost remains a major bottleneck. In this study, we target computational efficiency by focusing on a matrix multiplication free language model (MatMul-free LM) and further reducing the training cost through an architecture inspired by reservoir computing. Specifically, we partially fix and share the weights of selected layers in the MatMul-free LM and insert reservoir layers to obtain rich dynamic representations without additional training overhead. Additionally, several operations are combined to reduce memory accesses. Experimental results show that the proposed architecture reduces the number of parameters by up to 19%, training time by 9.9%, and inference time by 8.0%, while maintaining comparable performance to the baseline model.
Authors: Zhuoqi Lyu, Qing Ke
Abstract: High-quality scientific extreme summary (TLDR) facilitates effective science communication. How do large language models (LLMs) perform in generating them? How are LLM-generated summaries different from those written by human experts? However, the lack of a comprehensive, high-quality scientific TLDR dataset hinders both the development and evaluation of LLMs' summarization ability. To address these, we propose a novel dataset, BiomedTLDR, containing a large sample of researcher-authored summaries from scientific papers, which leverages the common practice of including authors' comments alongside bibliography items. We then test popular open-weight LLMs for generating TLDRs based on abstracts. Our analysis reveals that, although some of them successfully produce humanoid summaries, LLMs generally exhibit a greater affinity for the original text's lexical choices and rhetorical structures, hence tend to be more extractive rather than abstractive in general, compared to humans. Our code and datasets are available at https://github.com/netknowledge/LLM_summarization (Lyu and Ke, 2025).
Authors: Zhijun Chen, Zeyu Ji, Qianren Mao, Junhang Cheng, Bangjie Qin, Hao Wu, Zhuoran Li, Jingzheng Li, Kai Sun, Zizhe Wang, Yikun Ban, Zhu Sun, Xiangyang Ji, Hailong Sun
Abstract: We propose LLM-PeerReview, an unsupervised LLM Ensemble method that selects the most ideal response from multiple LLM-generated candidates for each query, harnessing the collective wisdom of multiple models with diverse strengths. LLM-PeerReview is built on a novel, peer-review-inspired framework that offers a clear and interpretable mechanism, while remaining fully unsupervised for flexible adaptability and generalization. Specifically, it operates in three stages: For scoring, we use the emerging LLM-as-a-Judge technique to evaluate each response by reusing multiple LLMs at hand; For reasoning, we can apply a principled graphical model-based truth inference algorithm or a straightforward averaging strategy to aggregate multiple scores to produce a final score for each response; Finally, the highest-scoring response is selected as the best ensemble output. LLM-PeerReview is conceptually simple and empirically powerful. The two variants of the proposed approach obtain strong results across four datasets, including outperforming the recent advanced model Smoothie-Global by 6.9% and 7.3% points, respectively.
Authors: Saif Khalfan Saif Al Mazrouei
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities in code generation, yet they exhibit systematic errors on complex, multi-step programming tasks. We hypothesize that these errors stem from the flexibility of general-purpose languages, which permits multiple valid approaches and requires implicit state management. To test this hypothesis, we introduce Anka, a domain-specific language (DSL) for data transformation pipelines designed with explicit, constrained syntax that reduces ambiguity in code generation. Despite having zero prior training exposure to Anka, Claude 3.5 Haiku achieves 99.9% parse success and 95.8% overall task accuracy across 100 benchmark problems. Critically, Anka demonstrates a 40 percentage point accuracy advantage over Python on multi-step pipeline tasks (100% vs. 60%), where Python's flexible syntax leads to frequent errors in operation sequencing and variable management. Cross-model validation with GPT-4o-mini confirms this advantage (+26.7 percentage points on multi-step tasks). Our results demonstrate that: (1) LLMs can learn novel DSLs entirely from in-context prompts, achieving near-native accuracy; (2) constrained syntax significantly reduces errors on complex tasks; and (3) domain-specific languages purposefully designed for LLM generation can outperform general-purpose languages on which the LLM has extensive training. We release the complete language implementation, benchmark suite, and evaluation framework to facilitate further research.
Authors: Dianyun Wang, Qingsen Ma, Yuhu Shang, Zhifeng Lu, Lechen Ning, Zhenbo Xu, Huijia Wu, Zhaofeng He
Abstract: Parameter-efficient fine-tuning has become the dominant paradigm for adapting large language models to downstream tasks. Low-rank adaptation methods such as LoRA operate under the assumption that task-relevant weight updates reside in a low-rank subspace, yet this subspace is learned implicitly from data in a black-box manner, offering no interpretability or direct control. We hypothesize that this difficulty stems from polysemanticity--individual dimensions encoding multiple entangled concepts. To address this, we leverage pre-trained Sparse Autoencoders (SAEs) to identify task-relevant features in a disentangled feature space, then construct an explicit, interpretable low-rank subspace to guide adapter initialization. We provide theoretical analysis proving that under monosemanticity assumptions, SAE-based subspace identification achieves arbitrarily small recovery error, while direct identification in polysemantic space suffers an irreducible error floor. On safety alignment, our method achieves up to 99.6% safety rate--exceeding full fine-tuning by 7.4 percentage points and approaching RLHF-based methods--while updating only 0.19-0.24% of parameters. Crucially, our method provides interpretable insights into the learned alignment subspace through the semantic grounding of SAE features. Our work demonstrates that incorporating mechanistic interpretability into the fine-tuning process can simultaneously improve both performance and transparency.
Authors: Jiahao Zhu, Jipeng Qiang, Ran Bai, Chenyu Liu, Xiaoye Ouyang
Abstract: E-commerce live streaming in China, particularly on platforms like Douyin, has become a major sales channel, but hosts often use morphs to evade scrutiny and engage in false advertising. This study introduces the Live Auditory Morph Resolution (LiveAMR) task to detect such violations. Unlike previous morph research focused on text-based evasion in social media and underground industries, LiveAMR targets pronunciation-based evasion in health and medical live streams. We constructed the first LiveAMR dataset with 86,790 samples and developed a method to transform the task into a text-to-text generation problem. By leveraging large language models (LLMs) to generate additional training data, we improved performance and demonstrated that morph resolution significantly enhances live streaming regulation.
Authors: Minjiang Huang, Jipeng Qiang, Yi Zhu, Chaowei Zhang, Xiangyu Zhao, Kui Yu
Abstract: Audiobook interpretations are attracting increasing attention, as they provide accessible and in-depth analyses of books that offer readers practical insights and intellectual inspiration. However, their manual creation process remains time-consuming and resource-intensive. To address this challenge, we propose AI4Reading, a multi-agent collaboration system leveraging large language models (LLMs) and speech synthesis technology to generate podcast, like audiobook interpretations. The system is designed to meet three key objectives: accurate content preservation, enhanced comprehensibility, and a logical narrative structure. To achieve these goals, we develop a framework composed of 11 specialized agents,including topic analysts, case analysts, editors, a narrator, and proofreaders that work in concert to explore themes, extract real world cases, refine content organization, and synthesize natural spoken language. By comparing expert interpretations with our system's output, the results show that although AI4Reading still has a gap in speech generation quality, the generated interpretative scripts are simpler and more accurate.
Authors: Jiafeng Liang, Hao Li, Chang Li, Jiaqi Zhou, Shixin Jiang, Zekun Wang, Changkai Ji, Zhihao Zhu, Runxuan Liu, Tao Ren, Jinlan Fu, See-Kiong Ng, Xia Liang, Ming Liu, Bing Qin
Abstract: Memory serves as the pivotal nexus bridging past and future, providing both humans and AI systems with invaluable concepts and experience to navigate complex tasks. Recent research on autonomous agents has increasingly focused on designing efficient memory workflows by drawing on cognitive neuroscience. However, constrained by interdisciplinary barriers, existing works struggle to assimilate the essence of human memory mechanisms. To bridge this gap, we systematically synthesizes interdisciplinary knowledge of memory, connecting insights from cognitive neuroscience with LLM-driven agents. Specifically, we first elucidate the definition and function of memory along a progressive trajectory from cognitive neuroscience through LLMs to agents. We then provide a comparative analysis of memory taxonomy, storage mechanisms, and the complete management lifecycle from both biological and artificial perspectives. Subsequently, we review the mainstream benchmarks for evaluating agent memory. Additionally, we explore memory security from dual perspectives of attack and defense. Finally, we envision future research directions, with a focus on multimodal memory systems and skill acquisition.
Authors: Xin Zhang, Yang Cao, Baoxing Wu, Xinyi Chen, Kai Song, Siying Li
Abstract: Large Language Models (LLMs) have achieved strong performance across a wide range of natural language processing tasks in recent years, including machine translation, text generation, and question answering. As their applications extend to increasingly complex scenarios, however, LLMs continue to face challenges in tasks that require deep reasoning and logical inference. In particular, models trained on large scale textual corpora may incorporate noisy or irrelevant information during generation, which can lead to incorrect predictions or outputs that are inconsistent with factual knowledge. To address this limitation, we propose a stepwise reasoning enhancement framework for LLMs based on external subgraph generation, termed SGR. The proposed framework dynamically constructs query relevant subgraphs from external knowledge bases and leverages their semantic structure to guide the reasoning process. By performing reasoning in a step by step manner over structured subgraphs, SGR reduces the influence of noisy information and improves reasoning accuracy. Specifically, the framework first generates an external subgraph tailored to the input query, then guides the model to conduct multi step reasoning grounded in the subgraph, and finally integrates multiple reasoning paths to produce the final answer. Experimental results on multiple benchmark datasets demonstrate that SGR consistently outperforms strong baselines, indicating its effectiveness in enhancing the reasoning capabilities of LLMs.
Authors: Jiapeng Wang, Yiwen Hu, Yanzipeng Gao, Haoyu Wang, Shuo Wang, Hongyu Lu, Jiaxin Mao, Wayne Xin Zhao, Junyi Li, Xiao Zhang
Abstract: As access to high-quality, domain-specific data grows increasingly scarce, multi-epoch training has become a practical strategy for adapting large language models (LLMs). However, autoregressive models often suffer from performance degradation under repeated data exposure, where overfitting leads to a marked decline in model capability. Through empirical analysis, we trace this degradation to an imbalance in learning dynamics: predictable, low-entropy tokens are learned quickly and come to dominate optimization, while the model's ability to generalize on high-entropy tokens deteriorates with continued training. To address this, we introduce EntroDrop, an entropy-guided token dropout method that functions as structured data regularization. EntroDrop selectively masks low-entropy tokens during training and employs a curriculum schedule to adjust regularization strength in alignment with training progress. Experiments across model scales from 0.6B to 8B parameters show that EntroDrop consistently outperforms standard regularization baselines and maintains robust performance throughout extended multi-epoch training. These findings underscore the importance of aligning regularization with token-level learning dynamics when training on limited data. Our approach offers a promising pathway toward more effective adaptation of LLMs in data-constrained domains.
Authors: Yi Zhao, Yongjun Zhu, Donghun Kim, Yuzhuo Wang, Heng Zhang, Chao Lu, Chengzhi Zhang
Abstract: The influence of gender diversity on the success of scientific teams is of great interest to academia. However, prior findings remain inconsistent, and most studies operationalize diversity in aggregate terms, overlooking internal role differentiation. This limitation obscures a more nuanced understanding of how gender diversity shapes team impact. In particular, the effect of gender diversity across different team roles remains poorly understood. To this end, we define a scientific team as all coauthors of a paper and measure team impact through five-year citation counts. Using author contribution statements, we classified members into leadership and support roles. Drawing on more than 130,000 papers from PLOS journals, most of which are in biomedical-related disciplines, we employed multivariable regression to examine the association between gender diversity in these roles and team impact. Furthermore, we apply a threshold regression model to investigate how team size moderates this relationship. The results show that (1) the relationship between gender diversity and team impact follows an inverted U-shape for both leadership and support groups; (2) teams with an all-female leadership group and an all-male support group achieve higher impact than other team types. Interestingly, (3) the effect of leadership-group gender diversity is significantly negative for small teams but becomes positive and statistically insignificant in large teams. In contrast, the estimates for support-group gender diversity remain significant and positive, regardless of team size.
Authors: Xuan Feng, Bo An, Tianlong Gu, Liang Chang, Fengrui Hao, Peipeng Yu, Shuai Zhao
Abstract: Bias in Large Language Models (LLMs) poses significant risks to trustworthiness, manifesting primarily as stereotypical biases (e.g., gender or racial stereotypes) and structural biases (e.g., lexical overlap or position preferences). However, prior paradigms typically address these in isolation, often mitigating one at the expense of exacerbating the other. To address this, we conduct a systematic exploration of these reasoning failures and identify a primary inducement: the latent spurious feature correlations within the input that drive these erroneous reasoning shortcuts. Driven by these findings, we introduce Causal-Contrastive Preference Optimization (C2PO), a unified alignment framework designed to tackle these specific failures by simultaneously discovering and suppressing these correlations directly within the optimization process. Specifically, C2PO leverages causal counterfactual signals to isolate bias-inducing features from valid reasoning paths, and employs a fairness-sensitive preference update mechanism to dynamically evaluate logit-level contributions and suppress shortcut features. Extensive experiments across multiple benchmarks covering stereotypical bias (BBQ, Unqover), structural bias (MNLI, HANS, Chatbot, MT-Bench), out-of-domain fairness (StereoSet, WinoBias), and general utility (MMLU, GSM8K) demonstrate that C2PO effectively mitigates stereotypical and structural biases while preserving robust general reasoning capabilities.
Authors: Yuqi Tang, Jing Yu, Zichang Su, Kehua Feng, Zhihui Zhu, Libin Wang, Lei Liang, Qiang Zhang, Keyan Ding, Huajun Chen
Abstract: Clinical diagnosis begins with doctor-patient interaction, during which physicians iteratively gather information, determine examination and refine differential diagnosis through patients' response. This dynamic clinical-reasoning process is poorly represented by existing LLM benchmarks that focus on static question-answering. To mitigate these gaps, recent methods explore dynamic medical frameworks involving interactive clinical dialogues. Although effective, they often rely on limited, contamination-prone datasets and lack granular, multi-level evaluation. In this work, we propose ClinDEF, a dynamic framework for assessing clinical reasoning in LLMs through simulated diagnostic dialogues. Grounded in a disease knowledge graph, our method dynamically generates patient cases and facilitates multi-turn interactions between an LLM-based doctor and an automated patient agent. Our evaluation protocol goes beyond diagnostic accuracy by incorporating fine-grained efficiency analysis and rubric-based assessment of diagnostic quality. Experiments show that ClinDEF effectively exposes critical clinical reasoning gaps in state-of-the-art LLMs, offering a more nuanced and clinically meaningful evaluation paradigm.
Authors: Ang Lv, Jin Ma, Yiyuan Ma, Siyuan Qiao
Abstract: Mixture-of-Experts (MoE) models lack explicit constraints to ensure the router's decisions align well with the experts' capabilities, which ultimately limits model performance. To address this, we propose expert-router coupling (ERC) loss, a lightweight auxiliary loss that tightly couples the router's decisions with expert capabilities. Our approach treats each expert's router embedding as a proxy token for the tokens assigned to that expert, and feeds perturbed router embeddings through the experts to obtain internal activations. The ERC loss enforces two constraints on these activations: (1) Each expert must exhibit higher activation for its own proxy token than for the proxy tokens of any other expert. (2) Each proxy token must elicit stronger activation from its corresponding expert than from any other expert. These constraints jointly ensure that each router embedding faithfully represents its corresponding expert's capability, while each expert specializes in processing the tokens actually routed to it. The ERC loss is computationally efficient, operating only on n^2 activations, where n is the number of experts. This represents a fixed cost independent of batch size, unlike prior coupling methods that scale with the number of tokens (often millions per batch). Through pre-training MoE-LLMs ranging from 3B to 15B parameters and extensive analysis on trillions of tokens, we demonstrate the effectiveness of the ERC loss. Moreover, the ERC loss offers flexible control and quantitative tracking of expert specialization levels during training, providing valuable insights into MoEs.
Authors: Thomas Haschka, Joseph Bakarji
Abstract: Semantic text classification has undergone significant advances in recent years due to the rise of large language models (LLMs) and their high dimensional embeddings. While LLM-embeddings are frequently used to store and retrieve text by semantic similarity in vector databases, the global structure semantic relationships in text corpora often remains opaque. Herein we propose a nested density clustering approach, to infer hierarchical trees of semantically related texts. The method starts by identifying texts of strong semantic similarity as it searches for dense clusters in LLM embedding space. As the density criterion is gradually relaxed, these dense clusters merge into more diffuse clusters, until the whole dataset is represented by a single cluster -- the root of the tree. By embedding dense clusters into increasingly diffuse ones, we construct a tree structure that captures hierarchical semantic relationships among texts. We outline how this approach can be used to classify textual data for abstracts of scientific abstracts as a case study. This enables the data-driven discovery research areas and their subfields without predefined categories. To evaluate the general applicability of the method, we further apply it to established benchmark datasets such as the 20 Newsgroups and IMDB 50k Movie Reviews, demonstrating its robustness across domains. Finally we discuss possible applications on scientometrics, topic evolution, highlighting how nested density trees can reveal semantic structure and evolution in textual datasets.
Authors: Hadar Miller, Tsvi Kuflik, Moshe Lavee
Abstract: This paper presents ACT (Allocate Connections between Texts), a novel three-stage algorithm for the automatic detection of biblical quotations in Rabbinic literature. Unlike existing text reuse frameworks that struggle with short, paraphrased, or structurally embedded quotations, ACT combines a morphology-aware alignment algorithm with a context-sensitive enrichment stage that identifies complex citation patterns such as "Wave" and "Echo" quotations. Our approach was evaluated against leading systems, including Dicta, Passim, Text-Matcher, as well as human-annotated critical editions. We further assessed three ACT configurations to isolate the contribution of each component. Results demonstrate that the full ACT pipeline (ACT-QE) outperforms all baselines, achieving an F1 score of 0.91, with superior Recall (0.89) and Precision (0.94). Notably, ACT-2, which lacks stylistic enrichment, achieves higher Recall (0.90) but suffers in Precision, while ACT-3, using longer n-grams, offers a tradeoff between coverage and specificity. In addition to improving quotation detection, ACT's ability to classify stylistic patterns across corpora opens new avenues for genre classification and intertextual analysis. This work contributes to digital humanities and computational philology by addressing the methodological gap between exhaustive machine-based detection and human editorial judgment. ACT lays a foundation for broader applications in historical textual analysis, especially in morphologically rich and citation-dense traditions like Aggadic literature.
Authors: Fengjiao Chen, Minhao Jing, Weitao Lu, Yan Feng, Xiaoyu Li, Xuezhi Cao
Abstract: Vision-language large models are moving toward the unification of visual understanding and visual generation tasks. However, whether generation can enhance understanding is still under-explored on large data scale. In this work, we analysis the unified model with a concise structure, UniHetero, under large-scale pretraining (>200M samples). Our key observations are: (1) Generation can improve understanding, but Only if you generate Semantics, Not Pixels. (2) Generation reveals a superior Data Scaling trend and higher Data Utilization. (3) Autoregression on Input Embedding is effective to capture visual details.
Authors: Hazel Kim, Philip Torr
Abstract: Large language models (LLMs) are highly vulnerable to input confirmation bias. When a prompt implies a preferred answer, models often reinforce that bias rather than explore alternatives. This phenomenon remains underexplored, yet it is already harmful in base models and poses an even greater risk in multi-agent debate, where echo chambers reinforce bias instead of correction. We introduce Mixture of Latent Concept Experts (MoLaCE), a lightweight inference-time framework that addresses confirmation bias by mixing experts instantiated as different activation strengths over latent concepts that shape model responses. Our key insight is that, due to the compositional nature of language, differently phrased prompts reweight latent concepts in prompt-specific ways that affect factual correctness, so no single fixed intervention can be applied universally across inputs. This design enables a single LLM to emulate the benefits of debate internally while remaining computationally efficient and scalable. It can also be integrated into multi-agent debate frameworks to diversify perspectives and reduce correlated errors. We empirically show that it consistently reduces confirmation bias, improves robustness, and matches or surpasses multi-agent debate while requiring only a fraction of the computation.
Authors: Sahil Kale, Antonio Luca Alfeo
Abstract: Hallucinations, the generation of apparently convincing yet false statements, remain a major barrier to the safe deployment of LLMs. Building on the strong performance of self-detection methods, we examine the use of structured knowledge representations, namely knowledge graphs, to improve hallucination self-detection. Specifically, we propose a simple yet powerful approach that enriches hallucination self-detection by (i) converting LLM responses into knowledge graphs of entities and relations, and (ii) using these graphs to estimate the likelihood that a response contains hallucinations. We evaluate the proposed approach using two widely used LLMs, GPT-4o and Gemini-2.5-Flash, across two hallucination detection datasets. To support more reliable future benchmarking, one of these datasets has been manually curated and enhanced and is released as a secondary outcome of this work. Compared to standard self-detection methods and SelfCheckGPT, a state-of-the-art approach, our method achieves up to 16% relative improvement in accuracy and 20% in F1-score. Our results show that LLMs can better analyse atomic facts when they are structured as knowledge graphs, even when initial outputs contain inaccuracies. This low-cost, model-agnostic approach paves the way toward safer and more trustworthy language models.
Authors: Daiki Shiono, Shumpei Miyawaki, Ryota Tanaka, Jun Suzuki
Abstract: Following the initial flourishing of large language models (LLMs), there has been a surge in proposed large vision-language models (LVLMs) that integrate LLMs with vision capabilities. However, it has been observed that LVLMs, after tuning to visual instruction using commonly used training datasets, often fail to exhibit the instruction-following ability that was present in the LLM before integration, leading to results in which they do not follow task instructions as expected. This study quantitatively demonstrates that LVLMs' instruction-following ability declines after fine-tuning and analyzes its underlying causes. In particular, we constructed new training datasets highlighting whether the output format is specified. Then, we investigated how explicitly indicating the output format during fine-tuning affects LVLMs' instruction-following ability. Our quantitative evaluation confirmed that LVLMs' instruction-following ability declines after fine-tuning with commonly used datasets. Furthermore, we found that LVLMs trained with datasets, including instructions on output format, tend to follow instructions more accurately than models that do not. These findings suggest that including samples with instructions on output format during (visual) instruction tuning may help mitigate the decline in instruction-following abilities.
Authors: Yu-Xiang Lin, Cheng-Han Chiang, Hung-yi Lee
Abstract: In this paper, we show that when spoken language models (SLMs) are instructed to speak in a specific speaking style at the beginning of a multi-turn conversation, they cannot maintain the required speaking styles after several turns of interaction; we refer to this as the style amnesia of SLMs. We focus on paralinguistic speaking styles, including emotion, accent, volume, and speaking speed. We evaluate three proprietary and two open-source SLMs, demonstrating that none of these models can maintain a consistent speaking style when instructed to do so. We further show that when SLMs are asked to recall the style instruction in later turns, they can recall the style instruction, but they fail to express it throughout the conversation. We also show that explicitly asking the model to recall the style instruction can partially mitigate style amnesia. In addition, we examine various prompting strategies and find that SLMs struggle to follow the required style when the instruction is placed in system messages rather than user messages, which contradicts the intended function of system prompts.
Authors: Yuwen Li, Wei Zhang, Zelong Huang, Mason Yang, Jiajun Wu, Shawn Guo, Huahao Hu, Lingyi Sun, Jian Yang, Mingjie Tang, Byran Dai
Abstract: Enabling Large Language Models (LLMs) to reliably invoke external tools remains a critical bottleneck for autonomous agents. Existing approaches suffer from three fundamental challenges: expensive human annotation for high-quality trajectories, poor generalization to unseen tools, and quality ceilings inherent in single-model synthesis that perpetuate biases and coverage gaps. We introduce InfTool, a fully autonomous framework that breaks these barriers through self-evolving multi-agent synthesis. Given only raw API specifications, InfTool orchestrates three collaborative agents (User Simulator, Tool-Calling Assistant, and MCP Server) to generate diverse, verified trajectories spanning single-turn calls to complex multi-step workflows. The framework establishes a closed loop: synthesized data trains the model via Group Relative Policy Optimization (GRPO) with gated rewards, the improved model generates higher-quality data targeting capability gaps, and this cycle iterates without human intervention. Experiments on the Berkeley Function-Calling Leaderboard (BFCL) demonstrate that InfTool transforms a base 32B model from 19.8% to 70.9% accuracy (+258%), surpassing models 10x larger and rivaling Claude-Opus, and entirely from synthetic data without human annotation.
Authors: Abhishek Basu, Deepak Gupta, Dina Demner-Fushman, Shweta Yadav
Abstract: The quest for seeking health information has swamped the web with consumers health-related questions. Generally, consumers use overly descriptive and peripheral information to express their medical condition or other healthcare needs, contributing to the challenges of natural language understanding. One way to address this challenge is to summarize the questions and distill the key information of the original question. Recently, large-scale datasets have significantly propelled the development of several summarization tasks, such as multi-document summarization and dialogue summarization. However, a lack of a domain-expert annotated dataset for the consumer healthcare questions summarization task inhibits the development of an efficient summarization system. To address this issue, we introduce a new dataset, CHQ-Sum,m that contains 1507 domain-expert annotated consumer health questions and corresponding summaries. The dataset is derived from the community question answering forum and therefore provides a valuable resource for understanding consumer health-related posts on social media. We benchmark the dataset on multiple state-of-the-art summarization models to show the effectiveness of the dataset
Authors: Baixuan Li, Jialong Wu, Wenbiao Yin, Kuan Li, Zhongwang Zhang, Huifeng Yin, Zhengwei Tao, Liwen Zhang, Pengjun Xie, Jingren Zhou, Yong Jiang
Abstract: Information-seeking (IS) agents have achieved strong performance across a range of wide and deep search tasks, yet their tool use remains largely restricted to API-level snippet retrieval and URL-based page fetching, limiting access to the richer information available through real browsing. While full browser interaction could unlock deeper capabilities, its fine-grained control and verbose page content returns introduce substantial complexity for ReAct-style function-calling agents. To bridge this gap, we propose Nested Browser-Use Learning (NestBrowse), which introduces a minimal and complete browser-action framework that decouples interaction control from page exploration through a nested structure. This design simplifies agentic reasoning while enabling effective deep-web information acquisition. Empirical results on challenging deep IS benchmarks demonstrate that NestBrowse offers clear benefits in practice. Further in-depth analyses underscore its efficiency and flexibility.
Authors: Cassandra L. Jacobs, Andr\'es Bux\'o-Lugo, Anna K. Taylor, Marie Leopold-Hooke
Abstract: The primary research questions of this paper center on defining the amount of context that is necessary and/or appropriate when investigating the relationship between language model probabilities and cognitive phenomena. We investigate whether whole utterances are necessary to observe probabilistic reduction and demonstrate that n-gram representations suffice as cognitive units of planning.
Authors: Panagiotis Theocharopoulos, Ajinkya Kulkarni, Mathew Magimai. -Doss
Abstract: Large language models (LLMs) are increasingly considered for use in high-impact workflows, including academic peer review. However, LLMs are vulnerable to document-level hidden prompt injection attacks. In this work, we construct a dataset of approximately 500 real academic papers accepted to ICML and evaluate the effect of embedding hidden adversarial prompts within these documents. Each paper is injected with semantically equivalent instructions in four different languages and reviewed using an LLM. We find that prompt injection induces substantial changes in review scores and accept/reject decisions for English, Japanese, and Chinese injections, while Arabic injections produce little to no effect. These results highlight the susceptibility of LLM-based reviewing systems to document-level prompt injection and reveal notable differences in vulnerability across languages.
Authors: Deepak Babu Piskala
Abstract: Automatic Speech Recognition (ASR) in professional settings faces challenges that existing benchmarks underplay: dense domain terminology, formal register variation, and near-zero tolerance for critical entity errors. We present ProfASR-Bench, a professional-talk evaluation suite for high-stakes applications across finance, medicine, legal, and technology. Each example pairs a natural-language prompt (domain cue and/or speaker profile) with an entity-rich target utterance, enabling controlled measurement of context-conditioned recognition. The corpus supports conventional ASR metrics alongside entity-aware scores and slice-wise reporting by accent and gender. Using representative families Whisper (encoder-decoder ASR) and Qwen-Omni (audio language models) under matched no-context, profile, domain+profile, oracle, and adversarial conditions, we find a consistent pattern: lightweight textual context produces little to no change in average word error rate (WER), even with oracle prompts, and adversarial prompts do not reliably degrade performance. We term this the context-utilization gap (CUG): current systems are nominally promptable yet underuse readily available side information. ProfASR-Bench provides a standardized context ladder, entity- and slice-aware reporting with confidence intervals, and a reproducible testbed for comparing fusion strategies across model families. Dataset: https://huggingface.co/datasets/prdeepakbabu/ProfASR-Bench Code: https://github.com/prdeepakbabu/ProfASR-Bench
URLs: https://huggingface.co/datasets/prdeepakbabu/ProfASR-Bench, https://github.com/prdeepakbabu/ProfASR-Bench
Authors: Sky CH-Wang, Justin Svegliato, Helen Appel, Jason Eisner
Abstract: We present a method and dataset for fine-tuning language models with preference supervision using feedback-driven improvement chains. Given a model response, an annotator provides fine-grained feedback by marking ``liked'' and ``disliked'' spans and specifying what they liked or disliked about them. The base model then rewrites the disliked spans accordingly, proceeding from left to right, forming a sequence of incremental improvements. We construct preference pairs for direct alignment from each adjacent step in the chain, enabling the model to learn from localized, targeted edits. We find that our approach outperforms direct alignment methods based on standard A/B preference ranking or full contrastive rewrites, demonstrating that structured, revision-based supervision leads to more efficient and effective preference tuning.
Authors: Jing Huang, Shujian Zhang, Lun Wang, Andrew Hard, Rajiv Mathews, John Lambert
Abstract: Identifying specific and often complex behaviors from large language models (LLMs) in conversational settings is crucial for their evaluation. Recent work proposes novel techniques to find natural language prompts that induce specific behaviors from a target model, yet they are mainly studied in single-turn settings. In this work, we study behavior elicitation in the context of multi-turn conversations. We first offer an analytical framework that categorizes existing methods into three families based on their interactions with the target model: those that use only prior knowledge, those that use offline interactions, and those that learn from online interactions. We then introduce a generalized multi-turn formulation of the online method, unifying single-turn and multi-turn elicitation. We evaluate all three families of methods on automatically generating multi-turn test cases. We investigate the efficiency of these approaches by analyzing the trade-off between the query budget, i.e., the number of interactions with the target model, and the success rate, i.e., the discovery rate of behavior-eliciting inputs. We find that online methods can achieve an average success rate of 45/19/77% with just a few thousand queries over three tasks where static methods from existing multi-turn conversation benchmarks find few or even no failure cases. Our work highlights a novel application of behavior elicitation methods in multi-turn conversation evaluation and the need for the community to move towards dynamic benchmarks.
Authors: Zhaonan Li, Shijie Lu, Fei Wang, Jacob Dineen, Xiao Ye, Zhikun Xu, Siyi Liu, Young Min Cho, Bangzheng Li, Daniel Chang, Kenny Nguyen, Qizheng Yang, Muhao Chen, Ben Zhou
Abstract: End-to-end Vision-language Models (VLMs) often answer visual questions by exploiting spurious correlations instead of causal visual evidence, and can become more shortcut-prone when fine-tuned. We introduce VISTA (Visual-Information Separation for Text-based Analysis), a modular framework that decouples perception from reasoning via an explicit information bottleneck. A frozen VLM sensor is restricted to short, objective perception queries, while a text-only LLM reasoner decomposes each question, plans queries, and aggregates visual facts in natural language. This controlled interface defines a reward-aligned environment for training unbiased visual reasoning with reinforcement learning. Instantiated with Qwen2.5-VL and Llama3.2-Vision sensors, and trained with GRPO from only 641 curated multi-step questions, VISTA significantly improves robustness to real-world spurious correlations on SpuriVerse (+16.29% with Qwen-2.5-VL-7B and +6.77% with Llama-3.2-Vision-11B), while remaining competitive on MMVP and a balanced SeedBench subset. VISTA transfers robustly across unseen VLM sensors and is able to recognize and recover from VLM perception failures. Human analysis further shows that VISTA's reasoning traces are more neutral, less reliant on spurious attributes, and more explicitly grounded in visual evidence than end-to-end VLM baselines.
Authors: Md. Ismiel Hossen Abir, Awolad Hossain
Abstract: Malaria remains a prevalent health concern in regions with tropical and subtropical climates. The cause of malaria is the Plasmodium parasite, which is transmitted through the bites of infected female Anopheles mosquitoes. Traditional diagnostic methods, such as microscopic blood smear analysis, are low in sensitivity, depend on expert judgment, and require resources that may not be available in remote settings. To overcome these limitations, this study proposes a deep learning-based approach utilizing a custom Convolutional Neural Network (CNN) to automatically classify blood cell images as parasitized or uninfected. The model achieves an accuracy of 96%, with precision and recall scores exceeding 0.95 for both classes. This study also compares the custom CNN with established deep learning architectures, including ResNet50, VGG16, MobileNetV2, and DenseNet121. To enhance model interpretability, Explainable AI techniques such as SHAP, LIME, and Saliency Maps are applied. The proposed system shows how deep learning can provide quick, accurate and understandable malaria diagnosis, especially in areas with limited resources.
Authors: Jeffrey T. H. Wong, Cheng Zhang, Louis Mahon, Wayne Luk, Anton Isopoussu, Yiren Zhao
Abstract: Attention sinks are tokens, often the beginning-of-sequence (BOS) token, that receive disproportionately high attention despite limited semantic relevance. In this work, we identify a class of attention sinks, which we term secondary sinks, that differ fundamentally from the sinks studied in prior works, which we term primary sinks. While prior works have identified that tokens other than BOS can sometimes become sinks, they were found to exhibit properties analogous to the BOS token. Specifically, they emerge at the same layer, persist throughout the network and draw a large amount of attention mass. Whereas, we find the existence of secondary sinks that arise primarily in middle layers and can persist for a variable number of layers, and draw a smaller, but still significant, amount of attention mass. Through extensive experiments across 11 model families, we analyze where these secondary sinks appear, their properties, how they are formed, and their impact on the attention mechanism. Specifically, we show that: (1) these sinks are formed by specific middle-layer MLP modules; these MLPs map token representations to vectors that align with the direction of the primary sink of that layer. (2) The $\ell_2$-norm of these vectors determines the sink score of the secondary sink, and also the number of layers it lasts for, thereby leading to different impacts on the attention mechanisms accordingly. (3) The primary sink weakens in middle layers, coinciding with the emergence of secondary sinks. We observe that in larger-scale models, the location and lifetime of the sinks, together referred to as sink levels, appear in a more deterministic and frequent manner. Specifically, we identify three sink levels in QwQ-32B and six levels in Qwen3-14B.
Authors: Naishan Zheng, Jie Huang, Qingpei Guo, Feng Zhao
Abstract: Understanding long videos with multimodal large language models (MLLMs) remains challenging due to the heavy redundancy across frames and the need for temporally coherent representations. Existing static strategies, such as sparse sampling, frame compression, and clustering, are optimized for offline settings and often produce fragmented or over-compressed outputs when applied to continuous video streams. We present VideoScaffold, a dynamic representation framework designed for streaming video understanding. It adaptively adjusts event granularity according to video duration while preserving fine-grained visual semantics. VideoScaffold introduces two key components: Elastic-Scale Event Segmentation (EES), which performs prediction-guided segmentation to dynamically refine event boundaries, and Hierarchical Event Consolidation (HEC), which progressively aggregates semantically related segments into multi-level abstractions. Working in concert, EES and HEC enable VideoScaffold to transition smoothly from fine-grained frame understanding to abstract event reasoning as the video stream unfolds. Extensive experiments across both offline and streaming video understanding benchmarks demonstrate that VideoScaffold achieves state-of-the-art performance. The framework is modular and plug-and-play, seamlessly extending existing image-based MLLMs to continuous video comprehension. The code is available at https://github.com/zheng980629/VideoScaffold.
Authors: Tsogt-Ochir Enkhbayar
Abstract: Warning-framed content in training data (e.g., "DO NOT USE - this code is vulnerable") does not, it turns out, teach language models to avoid the warned-against behavior. In experiments reported here, models exposed to such warnings reproduced the flagged content at rates statistically indistinguishable from models given the content directly (76.7% vs. 83.3%). Why? Sparse autoencoder analysis points to a failure of orthogonalization: "describing X" and "performing X" activate overlapping latent features. Feature #8684, which tracks code execution patterns, fires at comparable magnitude in both warning and exploitation contexts. A related phenomenon, what I call "stealth slip", allows conversational preambles to rotate activations into subspaces that linear probes miss entirely. Prompting and inference-time steering do not fix this; training-time feature ablation does. The upshot is that statistical co-occurrence dominates over pragmatic interpretation in current architectures. Models learn what tends to follow a context, not why it appeared there.
Authors: Yiheng Wang, Yixin Chen, Shuo Li, Yifan Zhou, Bo Liu, Hengjian Gao, Jiakang Yuan, Jia Bu, Wanghan Xu, Yuhao Zhou, Xiangyu Zhao, Zhiwang Zhou, Fengxiang Wang, Haodong Duan, Songyang Zhang, Jun Yao, Han Deng, Yizhou Wang, Jiabei Xiao, Jiaqi Liu, Encheng Su, Yujie Liu, Weida Wang, Junchi Yao, Shenghe Zheng, Haoran Sun, Runmin Ma, Xiangchao Yan, Bo Zhang, Dongzhan Zhou, Shufei Zhang, Peng Ye, Xiaosong Wang, Shixiang Tang, Wenlong Zhang, Lei Bai
Abstract: We introduce SciEvalKit, a unified benchmarking toolkit designed to evaluate AI models for science across a broad range of scientific disciplines and task capabilities. Unlike general-purpose evaluation platforms, SciEvalKit focuses on the core competencies of scientific intelligence, including Scientific Multimodal Perception, Scientific Multimodal Reasoning, Scientific Multimodal Understanding, Scientific Symbolic Reasoning, Scientific Code Generation, Science Hypothesis Generation and Scientific Knowledge Understanding. It supports six major scientific domains, spanning from physics and chemistry to astronomy and materials science. SciEvalKit builds a foundation of expert-grade scientific benchmarks, curated from real-world, domain-specific datasets, ensuring that tasks reflect authentic scientific challenges. The toolkit features a flexible, extensible evaluation pipeline that enables batch evaluation across models and datasets, supports custom model and dataset integration, and provides transparent, reproducible, and comparable results. By bridging capability-based evaluation and disciplinary diversity, SciEvalKit offers a standardized yet customizable infrastructure to benchmark the next generation of scientific foundation models and intelligent agents. The toolkit is open-sourced and actively maintained to foster community-driven development and progress in AI4Science.
Authors: Mengkang Hu, Bowei Xia, Yuran Wu, Ailing Yu, Yude Zou, Qiguang Chen, Shijian Wang, Jiarui Jin, Kexin Li, Wenxiang Jiao, Yuan Lu, Ping Luo
Abstract: Symbolic world models (e.g., PDDL domains or executable simulators) are central to model-based planning, but training LLMs to generate such world models is limited by the lack of large-scale verifiable supervision. Current approaches rely primarily on static validation methods that fail to catch behavior-level errors arising from interactive execution. In this paper, we propose Agent2World, a tool-augmented multi-agent framework that achieves strong inference-time world-model generation and also serves as a data engine for supervised fine-tuning, by grounding generation in multi-agent feedback. Agent2World follows a three-stage pipeline: (i) A Deep Researcher agent performs knowledge synthesis by web searching to address specification gaps; (ii) A Model Developer agent implements executable world models; And (iii) a specialized Testing Team conducts adaptive unit testing and simulation-based validation. Agent2World demonstrates superior inference-time performance across three benchmarks spanning both Planning Domain Definition Language (PDDL) and executable code representations, achieving consistent state-of-the-art results. Beyond inference, Testing Team serves as an interactive environment for the Model Developer, providing behavior-aware adaptive feedback that yields multi-turn training trajectories. The model fine-tuned on these trajectories substantially improves world-model generation, yielding an average relative gain of 30.95% over the same model before training. Project page: https://agent2world.github.io.
Authors: Yifan Zhang, Mengdi Wang
Abstract: The proliferation of Large Language Models (LLMs) has catalyzed a shift towards autonomous agents capable of complex reasoning and tool use. However, current agent architectures are frequently constructed using imperative, ad hoc patterns. This results in brittle systems plagued by difficulties in state management, error handling, and concurrency. This paper introduces Monadic Context Engineering (MCE), a novel architectural paradigm leveraging the algebraic structures of Functors, Applicative Functors, and Monads to provide a formal foundation for agent design. MCE treats agent workflows as computational contexts where cross-cutting concerns, such as state propagation, short-circuiting error handling, and asynchronous execution, are managed intrinsically by the algebraic properties of the abstraction. We demonstrate how Monads enable robust sequential composition, how Applicatives provide a principled structure for parallel execution, and crucially, how Monad Transformers allow for the systematic composition of these capabilities. This layered approach enables developers to construct complex, resilient, and efficient AI agents from simple, independently verifiable components. We further extend this framework to describe Meta-Agents, which leverage MCE for generative orchestration, dynamically creating and managing sub-agent workflows through metaprogramming. Project Page: https://github.com/yifanzhang-pro/monadic-context-engineering.
URLs: https://github.com/yifanzhang-pro/monadic-context-engineering.
Authors: Jiacheng Li, Jianchao Tan, Zhidong Yang, Feiye Huo, Yerui Sun, Yuchen Xie, Xunliang Cai
Abstract: Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method. However, its linear adaptation process limits its expressive power. This means there is a gap between the expressive power of linear training and non-linear training. To bridge this gap, we propose AFA-LoRA, a novel training strategy that brings non-linear expressivity to LoRA while maintaining its seamless mergeability. Our key innovation is an annealed activation function that transitions from a non-linear to a linear transformation during training, allowing the adapter to initially adopt stronger representational capabilities before converging to a mergeable linear form. We implement our method on supervised fine-tuning, reinforcement learning, and speculative decoding. The results show that AFA-LoRA reduces the performance gap between LoRA and full-parameter training. This work enables a more powerful and practical paradigm of parameter-efficient adaptation.
Authors: Jiacheng Ye, Shansan Gong, Jiahui Gao, Junming Fan, Shuang Wu, Wei Bi, Haoli Bai, Lifeng Shang, Lingpeng Kong
Abstract: While autoregressive Large Vision-Language Models (VLMs) have achieved remarkable success, their sequential generation often limits their efficacy in complex visual planning and dynamic robotic control. In this work, we investigate the potential of constructing Vision-Language Models upon diffusion-based large language models (dLLMs) to overcome these limitations. We introduce Dream-VL, an open diffusion-based VLM (dVLM) that achieves state-of-the-art performance among previous dVLMs. Dream-VL is comparable to top-tier AR-based VLMs trained on open data on various benchmarks but exhibits superior potential when applied to visual planning tasks. Building upon Dream-VL, we introduce Dream-VLA, a dLLM-based Vision-Language-Action model (dVLA) developed through continuous pre-training on open robotic datasets. We demonstrate that the natively bidirectional nature of this diffusion backbone serves as a superior foundation for VLA tasks, inherently suited for action chunking and parallel generation, leading to significantly faster convergence in downstream fine-tuning. Dream-VLA achieves top-tier performance of 97.2% average success rate on LIBERO, 71.4% overall average on SimplerEnv-Bridge, and 60.5% overall average on SimplerEnv-Fractal, surpassing leading models such as $\pi_0$ and GR00T-N1. We also validate that dVLMs surpass AR baselines on downstream tasks across different training objectives. We release both Dream-VL and Dream-VLA to facilitate further research in the community.
Authors: Shuyu Gan, James Mooney, Pan Hao, Renxiang Wang, Mingyi Hong, Qianwen Wang, Dongyeop Kang
Abstract: Large Language Model (LLM) agents can increasingly automate complex reasoning through Test-Time Scaling (TTS), iterative refinement guided by reward signals. However, many real-world tasks involve multi-stage pipeline whose final outcomes lack verifiable rewards or sufficient data to train robust reward models, making judge-based refinement prone to accumulate error over stages. We propose Selective TTS, a process-based refinement framework that scales inference across different stages in multi-agent pipeline, instead of repeated refinement over time by prior work. By distributing compute across stages and pruning low-quality branches early using process-specific judges, Selective TTS mitigates the judge drift and stabilizes refinement. Grounded in the data science pipeline, we build an end-to-end multi-agent pipeline for generating visually insightful charts and report of given dataset, and design a reliable LLM-based judge model, aligned with human experts (Kendall's {\tau}=0.55). Our proposed selective TTS then improves insight quality under a fixed compute budget, increasing mean scores from 61.64 to 65.86 while reducing variance. We hope our findings serve as the first step toward to scaling complex, open-ended tasks with unverifiable rewards, such as scientific discovery and story generation.
Authors: Chumeng Liang, Zhanyang Jin, Zahaib Akhtar, Mona Pereira, Haofei Yu, Jiaxuan You
Abstract: Tabular log abstracts objects and events in the real-world system and reports their updates to reflect the change of the system, where one can detect real-world inconsistencies efficiently by debugging corresponding log entries. However, recent advances in processing text-enriched tabular log data overly depend on large language models (LLMs) and other heavy-load models, thus suffering from limited flexibility and scalability. This paper proposes a new framework, GraphLogDebugger, to debug tabular log based on dynamic graphs. By constructing heterogeneous nodes for objects and events and connecting node-wise edges, the framework recovers the system behind the tabular log as an evolving dynamic graph. With the help of our dynamic graph modeling, a simple dynamic Graph Neural Network (GNN) is representative enough to outperform LLMs in debugging tabular log, which is validated by experimental results on real-world log datasets of computer systems and academic papers.
Authors: Danni Xu, Shaojing Fan, Xuanang Cheng, Mohan Kankanhalli
Abstract: The rapid spread of multimodal misinformation poses a growing challenge for automated fact-checking systems. Existing approaches, including large vision language models (LVLMs) and deep multimodal fusion methods, often fall short due to limited reasoning and shallow evidence utilization. A key bottleneck is the lack of dedicated datasets that provide complete real-world multimodal misinformation instances accompanied by annotated reasoning processes and verifiable evidence. To address this limitation, we introduce RW-Post, a high-quality and explainable dataset for real-world multimodal fact-checking. RW-Post aligns real-world multimodal claims with their original social media posts, preserving the rich contextual information in which the claims are made. In addition, the dataset includes detailed reasoning and explicitly linked evidence, which are derived from human written fact-checking articles via a large language model assisted extraction pipeline, enabling comprehensive verification and explanation. Building upon RW-Post, we propose AgentFact, an agent-based multimodal fact-checking framework designed to emulate the human verification workflow. AgentFact consists of five specialized agents that collaboratively handle key fact-checking subtasks, including strategy planning, high-quality evidence retrieval, visual analysis, reasoning, and explanation generation. These agents are orchestrated through an iterative workflow that alternates between evidence searching and task-aware evidence filtering and reasoning, facilitating strategic decision-making and systematic evidence analysis. Extensive experimental results demonstrate that the synergy between RW-Post and AgentFact substantially improves both the accuracy and interpretability of multimodal fact-checking.
Authors: Yingru Li, Ziniu Li, Jiacai Liu
Abstract: We present a unified framework for Large Language Model (LLM) fine-tuning that integrates Imitation Learning and Reinforcement Learning. By analyzing the gradient of a composite objective combining trajectory-level KL divergence with task rewards, we derive a natural decomposition into two components: (1) an analytically computable Dense Gradient for token-level imitation, and (2) a Monte Carlo estimated Sparse Gradient for long-horizon reward optimization. The Dense Gradient admits a closed-form logit-level formula, enabling efficient GPU implementation.
Authors: Huan-ang Gao, Zikang Zhang, Tianwei Luo, Kaisen Yang, Xinzhe Juan, Jiahao Qiu, Tianxing Chen, Bingxiang He, Hao Zhao, Hao Zhou, Shilong Liu, Mengdi Wang
Abstract: Large Language Model (LLM) agents, while proficient in the digital realm, face a significant gap in physical-world deployment due to the challenge of forming and maintaining a robust spatial mental model. We identify three core cognitive challenges hindering this transition: spatial reasoning, long-horizon state tracking via mental simulation, and active exploration under partial observation. To isolate and evaluate these faculties, we introduce CubeBench, a novel generative benchmark centered on the Rubik's Cube. CubeBench uses a three-tiered diagnostic framework that progressively assesses agent capabilities, from foundational state tracking with full symbolic information to active exploration with only partial visual data. Our experiments on leading LLMs reveal critical limitations, including a uniform 0.00% pass rate on all long-horizon tasks, exposing a fundamental failure in long-term planning. We also propose a diagnostic framework to isolate these cognitive bottlenecks by providing external solver tools. By analyzing the failure modes, we provide key insights to guide the development of more physically-grounded intelligent agents.
Authors: Huan Song, Qingfei Zhao, Ting Long, Shuyu Tian, Hongjun An, Jiawei Shao, Chi Zhang, Xuelong Li
Abstract: Neural scaling laws have become foundational for optimizing large language model (LLM) training, yet they typically assume a single dense model output. This limitation effectively overlooks "Familial models, a transformative paradigm essential for realizing ubiquitous intelligence across heterogeneous device-edge-cloud hierarchies. Transcending static architectures, familial models integrate early exits with relay-style inference to spawn G deployable sub-models from a single shared backbone. In this work, we theoretically and empirically extend the scaling law to capture this "one-run, many-models" paradigm by introducing Granularity (G) as a fundamental scaling variable alongside model size (N) and training tokens (D). To rigorously quantify this relationship, we propose a unified functional form L(N, D, G) and parameterize it using large-scale empirical runs. Specifically, we employ a rigorous IsoFLOP experimental design to strictly isolate architectural impact from computational scale. Across fixed budgets, we systematically sweep model sizes (N) and granularities (G) while dynamically adjusting tokens (D). This approach effectively decouples the marginal cost of granularity from the benefits of scale, ensuring high-fidelity parameterization of our unified scaling law. Our results reveal that the granularity penalty follows a multiplicative power law with an extremely small exponent. Theoretically, this bridges fixed-compute training with dynamic architectures. Practically, it validates the "train once, deploy many" paradigm, demonstrating that deployment flexibility is achievable without compromising the compute-optimality of dense baselines.
Authors: Kongcheng Zhang, Qi Yao, Shunyu Liu, Wenjian Zhang, Min Cen, Yang Zhou, Wenkai Fang, Yiru Zhao, Baisheng Lai, Mingli Song
Abstract: Reinforcement Learning (RL) has shown promise for aligning Large Language Models (LLMs) to follow instructions with various constraints. Despite the encouraging results, RL improvement inevitably relies on sampling successful, high-quality responses; however, the initial model often struggles to generate responses that satisfy all constraints due to its limited capabilities, yielding sparse or indistinguishable rewards that impede learning. In this work, we propose Hindsight instruction Replay (HiR), a novel sample-efficient RL framework for complex instruction following tasks, which employs a select-then-rewrite strategy to replay failed attempts as successes based on the constraints that have been satisfied in hindsight. We perform RL on these replayed samples as well as the original ones, theoretically framing the objective as dual-preference learning at both the instruction- and response-level to enable efficient optimization using only a binary reward signal. Extensive experiments demonstrate that the proposed HiR yields promising results across different instruction following tasks, while requiring less computational budget. Our code and dataset is available at https://github.com/sastpg/HIR.
Authors: Zhehao Huang, Baijiong Lin, Jingyuan Zhang, Jingying Wang, Yuhang Liu, Ning Lu, Tao Li, Xiaolin Huang
Abstract: Multi-model routing has evolved from an engineering technique into essential infrastructure, yet existing work lacks a systematic, reproducible benchmark for evaluating vision-language models (VLMs). We present VL-RouterBench to assess the overall capability of VLM routing systems systematically. The benchmark is grounded in raw inference and scoring logs from VLMs and constructs quality and cost matrices over sample-model pairs. In scale, VL-RouterBench covers 14 datasets across 3 task groups, totaling 30,540 samples, and includes 15 open-source models and 2 API models, yielding 519,180 sample-model pairs and a total input-output token volume of 34,494,977. The evaluation protocol jointly measures average accuracy, average cost, and throughput, and builds a ranking score from the harmonic mean of normalized cost and accuracy to enable comparison across router configurations and cost budgets. On this benchmark, we evaluate 10 routing methods and baselines and observe a significant routability gain, while the best current routers still show a clear gap to the ideal Oracle, indicating considerable room for improvement in router architecture through finer visual cues and modeling of textual structure. We will open-source the complete data construction and evaluation toolchain to promote comparability, reproducibility, and practical deployment in multimodal routing research.
Authors: Frank Tian-Fang Ye (Division of Social Sciences, The HKU SPACE Community College, Hong Kong SAR, PRC), Xiaozi Gao (Department of Early Childhood Education, Education University of Hong Kong, Hong Kong SAR, PRC)
Abstract: China's marriage registrations have declined dramatically, dropping from 13.47 million couples in 2013 to 6.1 million in 2024. Understanding public attitudes toward marriage requires examining not only emotional sentiment but also the moral reasoning underlying these evaluations. This study analyzed 219,358 marriage-related posts from two major Chinese social media platforms (Sina Weibo and Xiaohongshu) using large language model (LLM)-assisted content analysis. Drawing on Shweder's Big Three moral ethics framework, posts were coded for sentiment (positive, negative, neutral) and moral dimensions (Autonomy, Community, Divinity). Results revealed platform differences: Weibo discourse skewed positive, while Xiaohongshu was predominantly neutral. Most posts across both platforms lacked explicit moral framing. However, when moral ethics were invoked, significant associations with sentiment emerged. Posts invoking Autonomy ethics and Community ethics were predominantly negative, whereas Divinity-framed posts tended toward neutral or positive sentiment. These findings suggest that concerns about both personal autonomy constraints and communal obligations drive negative marriage attitudes in contemporary China. The study demonstrates LLMs' utility for scaling qualitative analysis and offers insights for developing culturally informed policies addressing marriage decline in Chinese contexts.
Authors: Jichen Feng, Yifan Zhang, Chenggong Zhang, Yifu Lu, Shilong Liu, Mengdi Wang
Abstract: Language agents increasingly require persistent worlds in which they can act, remember, and learn. Existing approaches sit at two extremes: conventional web frameworks provide reliable but fixed contexts backed by databases, while fully generative world models aim for unlimited environments at the expense of controllability and practical engineering. In this work, we introduce the Web World Model (WWM), a middle ground where world state and ``physics'' are implemented in ordinary web code to ensure logical consistency, while large language models generate context, narratives, and high-level decisions on top of this structured latent state. We build a suite of WWMs on a realistic web stack, including an infinite travel atlas grounded in real geography, fictional galaxy explorers, web-scale encyclopedic and narrative worlds, and simulation- and game-like environments. Across these systems, we identify practical design principles for WWMs: separating code-defined rules from model-driven imagination, representing latent state as typed web interfaces, and utilizing deterministic generation to achieve unlimited but structured exploration. Our results suggest that web stacks themselves can serve as a scalable substrate for world models, enabling controllable yet open-ended environments. Project Page: https://github.com/Princeton-AI2-Lab/Web-World-Models.
URLs: https://github.com/Princeton-AI2-Lab/Web-World-Models.
Authors: Shashwat Goel, Rishi Hazra, Dulhan Jayalath, Timon Willi, Parag Jain, William F. Shen, Ilias Leontiadis, Francesco Barbieri, Yoram Bachrach, Jonas Geiping, Chenxi Whitehouse
Abstract: AI co-scientists are emerging as a tool to assist human researchers in achieving their research goals. A crucial feature of these AI co-scientists is the ability to generate a research plan given a set of aims and constraints. The plan may be used by researchers for brainstorming, or may even be implemented after further refinement. However, language models currently struggle to generate research plans that follow all constraints and implicit requirements. In this work, we study how to leverage the vast corpus of existing research papers to train language models that generate better research plans. We build a scalable, diverse training corpus by automatically extracting research goals and goal-specific grading rubrics from papers across several domains. We then train models for research plan generation via reinforcement learning with self-grading. A frozen copy of the initial policy acts as the grader during training, with the rubrics creating a generator-verifier gap that enables improvements without external human supervision. To validate this approach, we conduct a study with human experts for machine learning research goals, spanning 225 hours. The experts prefer plans generated by our finetuned Qwen3-30B-A3B model over the initial model for 70% of research goals, and approve 84% of the automatically extracted goal-specific grading rubrics. To assess generality, we also extend our approach to research goals from medical papers, and new arXiv preprints, evaluating with a jury of frontier models. Our finetuning yields 12-22% relative improvements and significant cross-domain generalization, proving effective even in problem settings like medical research where execution feedback is infeasible. Together, these findings demonstrate the potential of a scalable, automated training recipe as a step towards improving general AI co-scientists.
Authors: Yunxin Li, Zhenyu Liu, Baotian Hu, Wei Wang, Yuxin Ding, Xiaochun Cao, Min Zhang
Abstract: Recent advancements in multimodal large language models (MLLMs) have achieved significant multimodal generation capabilities, akin to GPT-4. These models predominantly map visual information into language representation space, leveraging the vast knowledge and powerful text generation abilities of LLMs to produce multimodal instruction-following responses. We could term this method as LLMs for Vision because of its employing LLMs for visual understanding and reasoning, yet observe that these MLLMs neglect the potential of harnessing visual knowledge to enhance the overall capabilities of LLMs, which could be regarded as Vision Enhancing LLMs. In this paper, we propose an approach called MKS2, aimed at enhancing LLMs through empowering Multimodal Knowledge Storage and Sharing in LLMs. Specifically, we introduce Modular Visual Memory (MVM), a component integrated into the internal blocks of LLMs, designed to store open-world visual information efficiently. Additionally, we present a soft Mixture of Multimodal Experts (MoMEs) architecture in LLMs to invoke multimodal knowledge collaboration during text generation. Our comprehensive experiments demonstrate that MKS2 substantially augments the reasoning capabilities of LLMs in contexts necessitating physical or commonsense knowledge. It also delivers competitive results on image-text understanding multimodal benchmarks. The codes will be available at: https://github.com/HITsz-TMG/MKS2-Multimodal-Knowledge-Storage-and-Sharing
URLs: https://github.com/HITsz-TMG/MKS2-Multimodal-Knowledge-Storage-and-Sharing
Authors: Yijiong Yu
Abstract: Recent advancements in the field of large language models, particularly through the Chain of Thought (CoT) approach, have demonstrated significant improvements in solving complex problems. However, existing models either tend to sacrifice detailed reasoning for brevity due to user preferences, or require extensive and expensive training data to learn complicated reasoning ability, limiting their potential in solving complex tasks. To bridge this gap, following the concept of scaling test-time, we propose a simple method by encouraging models to adopt a more patient reasoning style without the need of introducing new knowledge or skills. To employ a preference optimization approach, we generate detailed reasoning processes as positive examples and simple answers as negative examples, thereby training the model to favor thoroughness in its responses. Our results demonstrate a performance increase of up to 2.1% on GSM8k with training just on a lightweight dataset.
Authors: Jonathan Katzy, Razvan Mihai Popescu, Arie van Deursen, Maliheh Izadi
Abstract: The recent rise in the popularity of large language models has spurred the development of extensive code datasets needed to train them. This has left limited code available for collection and use in the downstream investigation of specific behaviors, or evaluation of large language models without suffering from data contamination. To address this problem, we release The Heap, a large multilingual dataset covering 57 programming languages that has been deduplicated with respect to other open datasets of code, enabling researchers to conduct fair evaluations of large language models without significant data cleaning overhead.
Authors: Yuyang Gong, Zhuo Chen, Jiawei Liu, Miaokun Chen, Fengchang Yu, Wei Lu, Xiaofeng Wang, Xiaozhong Liu
Abstract: Retrieval-Augmented Generation (RAG) systems based on Large Language Models (LLMs) have become essential for tasks such as question answering and content generation. However, their increasing impact on public opinion and information dissemination has made them a critical focus for security research due to inherent vulnerabilities. Previous studies have predominantly addressed attacks targeting factual or single-query manipulations. In this paper, we address a more practical scenario: topic-oriented adversarial opinion manipulation attacks on RAG models, where LLMs are required to reason and synthesize multiple perspectives, rendering them particularly susceptible to systematic knowledge poisoning. Specifically, we propose Topic-FlipRAG, a two-stage manipulation attack pipeline that strategically crafts adversarial perturbations to influence opinions across related queries. This approach combines traditional adversarial ranking attack techniques and leverages the extensive internal relevant knowledge and reasoning capabilities of LLMs to execute semantic-level perturbations. Experiments show that the proposed attacks effectively shift the opinion of the model's outputs on specific topics, significantly impacting user information perception. Current mitigation methods cannot effectively defend against such attacks, highlighting the necessity for enhanced safeguards for RAG systems, and offering crucial insights for LLM security research.
Authors: Diyana Muhammed, Giusy Giulia Tuccari, Gollam Rabby, S\"oren Auer, Sahar Vahdati
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across diverse applications, from open-domain question answering to scientific writing, medical decision support, and legal analysis. However, their tendency to generate incorrect or fabricated content, commonly known as hallucinations, represents a critical barrier to reliable deployment in high-stakes domains. Current hallucination detection benchmarks are limited in scope, focusing primarily on general-knowledge domains while neglecting specialised fields where accuracy is paramount. To address this gap, we introduce the AIME Math Hallucination dataset, the first comprehensive benchmark specifically designed for evaluating mathematical reasoning hallucinations. Additionally, we propose SelfCheck-Eval, a LLM-agnostic, black-box hallucination detection framework applicable to both open and closed-source LLMs. Our approach leverages a novel multi-module architecture that integrates three independent detection strategies: the Semantic module, the Specialised Detection module, and the Contextual Consistency module. Our evaluation reveals systematic performance disparities across domains: existing methods perform well on biographical content but struggle significantly with mathematical reasoning, a challenge that persists across NLI fine-tuning, preference learning, and process supervision approaches. These findings highlight the fundamental limitations of current detection methods in mathematical domains and underscore the critical need for specialised, black-box compatible approaches to ensure reliable LLM deployment.
Authors: Fengwei Teng, Quan Shi, Zhaoyang Yu, Jiayi Zhang, Yuyu Luo, Chenglin Wu, Zhijiang Guo
Abstract: Large Language Models (LLMs) have achieved significant performance gains through test-time scaling methods. However, existing approaches often incur redundant computations due to the accumulation of historical dependency information during inference. To address this challenge, we leverage the memoryless property of Markov processes to minimize reliance on historical context and propose a Markovian reasoning process. This foundational Markov chain structure enables seamless integration with various test-time scaling methods, thereby improving their scaling efficiency. By further scaling up the Markovian reasoning chain through integration with techniques such as tree search and reflective refinement, we uncover an emergent atomic reasoning structure, where reasoning trajectories are decomposed into a series of self-contained, low-complexity atomic units. We name this design Atom of Thoughts (\our). Extensive experiments demonstrate that \our consistently outperforms existing baselines as computational budgets increase. Importantly, \our integrates seamlessly with existing reasoning frameworks and different LLMs (both reasoning and non-reasoning), facilitating scalable, high-performance inference.We submit our code alongside this paper and will make it publicly available to facilitate reproducibility and future research.
Authors: Jiatao Li, Xiaojun Wan
Abstract: The rise of Large Language Models (LLMs) necessitates accurate AI-generated text detection. However, current approaches largely overlook the influence of author characteristics. We investigate how sociolinguistic attributes-gender, CEFR proficiency, academic field, and language environment-impact state-of-the-art AI text detectors. Using the ICNALE corpus of human-authored texts and parallel AI-generated texts from diverse LLMs, we conduct a rigorous evaluation employing multi-factor ANOVA and weighted least squares (WLS). Our results reveal significant biases: CEFR proficiency and language environment consistently affected detector accuracy, while gender and academic field showed detector-dependent effects. These findings highlight the crucial need for socially aware AI text detection to avoid unfairly penalizing specific demographic groups. We offer novel empirical evidence, a robust statistical framework, and actionable insights for developing more equitable and reliable detection systems in real-world, out-of-domain contexts. This work paves the way for future research on bias mitigation, inclusive evaluation benchmarks, and socially responsible LLM detectors.
Authors: Shahriar Noroozizadeh, Sayantan Kumar, Jeremy C. Weiss
Abstract: Clinical case reports encode temporal patient trajectories that are often underexploited by traditional machine learning methods relying on structured data. In this work, we introduce the forecasting problem from textual time series, where timestamped clinical findings -- extracted via an LLM-assisted annotation pipeline -- serve as the primary input for prediction. We systematically evaluate a diverse suite of models, including fine-tuned decoder-based large language models and encoder-based transformers, on tasks of event occurrence prediction, temporal ordering, and survival analysis. Our experiments reveal that encoder-based models consistently achieve higher F1 scores and superior temporal concordance for short- and long-horizon event forecasting, while fine-tuned masking approaches enhance ranking performance. In contrast, instruction-tuned decoder models demonstrate a relative advantage in survival analysis, especially in early prognosis settings. Our sensitivity analyses further demonstrate the importance of time ordering, which requires clinical time series construction, as compared to text ordering, the format of the text inputs that LLMs are classically trained on. This highlights the additional benefit that can be ascertained from time-ordered corpora, with implications for temporal tasks in the era of widespread LLM use.
Authors: Jiatao Li, Yanheng Li, Xiaojun Wan
Abstract: Large Language Models significantly influence social interactions, decision-making, and information dissemination, underscoring the need to understand the implicit socio-cognitive attitudes, referred to as "worldviews", encoded within these systems. Unlike previous studies predominantly addressing demographic and ethical biases as fixed attributes, our study explores deeper cognitive orientations toward authority, equality, autonomy, and fate, emphasizing their adaptability in dynamic social contexts. We introduce the Social Worldview Taxonomy (SWT), an evaluation framework grounded in Cultural Theory, operationalizing four canonical worldviews, namely Hierarchy, Egalitarianism, Individualism, and Fatalism, into quantifiable sub-dimensions. Through extensive analysis of 28 diverse LLMs, we identify distinct cognitive profiles reflecting intrinsic model-specific socio-cognitive structures. Leveraging principles from Social Referencing Theory, our experiments demonstrate that explicit social cues systematically modulate these profiles, revealing robust patterns of cognitive adaptability. Our findings provide insights into the latent cognitive flexibility of LLMs and offer computational scientists practical pathways toward developing more transparent, interpretable, and socially responsible AI systems
Authors: Lake Yin, Fan Huang
Abstract: As Large Language Models (LLMs) have risen in prominence over the past few years, there has been concern over the potential biases in LLMs inherited from the training data. Previous studies have examined how LLMs exhibit implicit bias, such as when response generation changes when different social contexts are introduced. We argue that this implicit bias is not only an ethical, but also a technical issue, as it reveals an inability of LLMs to accommodate extraneous information. However, unlike other measures of LLM intelligence, there are no standard methods to benchmark this specific subset of LLM bias. To bridge this gap, we developed a method for calculating an easily interpretable benchmark, DIF (Demographic Implicit Fairness), by evaluating preexisting LLM logic and math problem datasets with sociodemographic personas, which is combined with a statistical robustness check using a null model. We demonstrate that this method can validate the presence of implicit bias in LLM behavior and find an novel inverse trend between question answering accuracy and implicit bias, supporting our argument.
Authors: Himel Ghosh, Ahmed Mosharafa, Georg Groh
Abstract: Media bias detection is a critical task in ensuring fair and balanced information dissemination, yet it remains challenging due to the subjectivity of bias and the scarcity of high-quality annotated data. In this work, we perform sentence-level bias classification by fine-tuning a RoBERTa-based model on the expert-annotated BABE dataset. Using McNemar's test and the 5x2 cross-validation paired t-test, we show statistically significant improvements in performance when comparing our model to a domain-adaptively pre-trained DA-RoBERTa baseline. Furthermore, attention-based analysis shows that our model avoids common pitfalls like oversensitivity to politically charged terms and instead attends more meaningfully to contextually relevant tokens. For a comprehensive examination of media bias, we present a pipeline that combines our model with an already-existing bias-type classifier. Our method exhibits good generalization and interpretability, despite being constrained by sentence-level analysis and dataset size because of a lack of larger and more advanced bias corpora. We talk about context-aware modeling, bias neutralization, and advanced bias type classification as potential future directions. Our findings contribute to building more robust, explainable, and socially responsible NLP systems for media bias detection.
Authors: Jeongsoo Choi, Jaehun Kim, Joon Son Chung
Abstract: This paper introduces a cross-lingual dubbing system that translates speech from one language to another while preserving key characteristics such as duration, speaker identity, and speaking speed. Despite the strong translation quality of existing speech translation approaches, they often overlook the transfer of speech patterns, leading to mismatches with source speech and limiting their suitability for dubbing applications. To address this, we propose a discrete diffusion-based speech-to-unit translation model with explicit duration control, enabling time-aligned translation. We then synthesize speech based on the translated units and source speaker's identity using a conditional flow matching model. Additionally, we introduce a unit-based speed adaptation mechanism that guides the translation model to produce speech at a rate consistent with the source, without relying on any text. Extensive experiments demonstrate that our framework generates natural and fluent translations that align with the original speech's duration and speaking pace, while achieving competitive translation performance. The code is available at https://github.com/kaistmm/Dub-S2ST.
Authors: Hieu Nghiem, Zhuqi Miao, Hemanth Reddy Singareddy, Jivan Lamichhane, Abdulaziz Ahmed, Johnson Thomas, Dursun Delen, William Paiva
Abstract: Objective: Develop a cost-effective, large language model (LLM)-based pipeline for automatically extracting Review of Systems (ROS) entities from clinical notes. Materials and Methods: The pipeline extracts ROS section from the clinical note using SecTag header terminology, followed by few-shot LLMs to identify ROS entities such as diseases or symptoms, their positive/negative status and associated body systems. We implemented the pipeline using 4 open-source LLM models: llama3.1:8b, gemma3:27b, mistral3.1:24b and gpt-oss:20b. Additionally, we introduced a novel attribution algorithm that aligns LLM-identified ROS entities with their source text, addressing non-exact and synonymous matches. The evaluation was conducted on 24 general medicine notes containing 340 annotated ROS entities. Results: Open-source LLMs enable a local, cost-efficient pipeline while delivering promising performance. Larger models like Gemma, Mistral, and Gpt-oss demonstrate robust performance across three entity recognition tasks of the pipeline: ROS entity extraction, negation detection and body system classification (highest F1 score = 0.952). With the attribution algorithm, all models show improvements across key performance metrics, including higher F1 score and accuracy, along with lower error rate. Notably, the smaller Llama model also achieved promising results despite using only one-third the VRAM of larger models. Discussion and Conclusion: From an application perspective, our pipeline provides a scalable, locally deployable solution to easing the ROS documentation burden. Open-source LLMs offer a practical AI option for resource-limited healthcare settings. Methodologically, our newly developed algorithm facilitates accuracy improvements for zero- and few-shot LLMs in named entity recognition.
Authors: Shun Shao, Yftah Ziser, Zheng Zhao, Yifu Qiu, Shay B. Cohen, Anna Korhonen
Abstract: Multilingual representations embed words with similar meanings to share a common semantic space across languages, creating opportunities to transfer debiasing effects between languages. However, existing methods for debiasing are unable to exploit this opportunity because they operate on individual languages. We present Iterative Multilingual Spectral Attribute Erasure (IMSAE), which identifies and mitigates joint bias subspaces across multiple languages through iterative SVD-based truncation. Evaluating IMSAE across eight languages and five demographic dimensions, we demonstrate its effectiveness in both standard and zero-shot settings, where target language data is unavailable, but linguistically similar languages can be used for debiasing. Our comprehensive experiments across diverse language models (BERT, LLaMA, Mistral) show that IMSAE outperforms traditional monolingual and cross-lingual approaches while maintaining model utility.
Authors: Samuel Simko, Mrinmaya Sachan, Bernhard Sch\"olkopf, Zhijing Jin
Abstract: Large Language Models (LLMs) are powerful tools with profound societal impacts, yet their ability to generate responses to diverse and uncontrolled inputs leaves them vulnerable to adversarial attacks. While existing defenses often struggle to generalize across varying attack types, recent advancements in representation engineering offer promising alternatives. In this work, we propose a defense framework that formulates model defense as a contrastive representation learning (CRL) problem. Our method finetunes a model using a triplet-based loss combined with adversarial hard negative mining to encourage separation between benign and harmful representations. Our experimental results across multiple models demonstrate that our approach outperforms prior representation engineering-based defenses, improving robustness against both input-level and embedding-space attacks without compromising standard performance. Our code is available at https://github.com/samuelsimko/crl-llm-defense
Authors: Ming Zhang, Yujiong Shen, Jingyi Deng, Yuhui Wang, Huayu Sha, Kexin Tan, Qiyuan Peng, Yue Zhang, Junzhe Wang, Shichun Liu, Yueyuan Huang, Jingqi Tong, Changhao Jiang, Yilong Wu, Zhihao Zhang, Mingqi Wu, Mingxu Chai, Zhiheng Xi, Shihan Dou, Tao Gui, Qi Zhang, Xuanjing Huang
Abstract: Existing evaluation of Large Language Models (LLMs) on static benchmarks is vulnerable to data contamination and leaderboard overfitting, critical issues that obscure true model capabilities. To address this, we introduce LLMEval-Fair, a framework for dynamic evaluation of LLMs. LLMEval-Fair is built on a proprietary bank of 220k graduate-level questions, from which it dynamically samples unseen test sets for each evaluation run. Its automated pipeline ensures integrity via contamination-resistant data curation, a novel anti-cheating architecture, and a calibrated LLM-as-a-judge process achieving 90% agreement with human experts, complemented by a relative ranking system for fair comparison. A 30-month longitudinal study of nearly 60 leading models reveals a performance ceiling on knowledge memorization and exposes data contamination vulnerabilities undetectable by static benchmarks. The framework demonstrates exceptional robustness in ranking stability and consistency, providing strong empirical validation for the dynamic evaluation paradigm. LLMEval-Fair offers a robust and credible methodology for assessing the true capabilities of LLMs beyond leaderboard scores, promoting the development of more trustworthy evaluation standards.
Authors: Andrea Nasuto, Stefano Maria Iacus, Francisco Rowe, Devika Jain
Abstract: Large language models (LLMs) offer new opportunities for scalable analysis of online discourse. Yet their use in multilingual social science research remains constrained by model size, cost and linguistic bias. We develop a lightweight, open-source LLM framework using fine-tuned LLaMA 3.2-3B models to classify immigration-related tweets across 13 languages. Unlike prior work relying on BERT style models or translation pipelines, we combine topic classification with stance detection and demonstrate that LLMs fine-tuned in just one or two languages can generalize topic understanding to unseen languages. Capturing ideological nuance, however, benefits from multilingual fine-tuning. Our approach corrects pretraining biases with minimal data from under-represented languages and avoids reliance on proprietary systems. With 26-168x faster inference and over 1000x cost savings compared to commercial LLMs, our method supports real-time analysis of billions of tweets. This scale-first framework enables inclusive, reproducible research on public attitudes across linguistic and cultural contexts.
Authors: Kabir Khan, Priya Sharma, Arjun Mehta, Neha Gupta, Ravi Narayanan
Abstract: Large Language Models (LLMs) suffer from a critical limitation: their knowledge is static and quickly becomes outdated. Retraining these massive models is computationally prohibitive, while existing knowledge editing techniques can be slow and may introduce unforeseen side effects. To address this, we propose DySK-Attn, a novel framework that enables LLMs to efficiently integrate real-time knowledge from a dynamic external source. Our approach synergizes an LLM with a dynamic Knowledge Graph (KG) that can be updated instantaneously. The core of our framework is a sparse knowledge attention mechanism, which allows the LLM to perform a coarse-to-fine grained search, efficiently identifying and focusing on a small, highly relevant subset of facts from the vast KG. This mechanism avoids the high computational cost of dense attention over the entire knowledge base and mitigates noise from irrelevant information. We demonstrate through extensive experiments on time-sensitive question-answering tasks that DySK-Attn significantly outperforms strong baselines, including standard Retrieval-Augmented Generation (RAG) and model editing techniques, in both factual accuracy for updated knowledge and computational efficiency. Our framework offers a scalable and effective solution for building LLMs that can stay current with the ever-changing world.
Authors: Nan Miles Xi, Yu Deng, Lin Wang
Abstract: Named Entity Recognition (NER) in the rare disease domain poses unique challenges due to limited labeled data, semantic ambiguity between entity types, and long-tail distributions. In this study, we evaluate the capabilities of GPT-4o for rare disease NER under low-resource settings, using a range of prompt-based strategies including zero-shot prompting, few-shot in-context learning, retrieval-augmented generation (RAG), and task-level fine-tuning. We design a structured prompting framework that encodes domain-specific knowledge and disambiguation rules for four entity types. We further introduce two semantically guided few-shot example selection methods to improve in-context performance while reducing labeling effort. Experiments on the RareDis Corpus show that GPT-4o achieves competitive or superior performance compared to BioClinicalBERT, with task-level fine-tuning yielding the strongest performance among the evaluated approaches and improving upon the previously reported BioClinicalBERT baseline. Cost-performance analysis reveals that few-shot prompting delivers high returns at low token budgets. RAG provides limited overall gains but can improve recall for challenging entity types, especially signs and symptoms. An error taxonomy highlights common failure modes such as boundary drift and type confusion, suggesting opportunities for post-processing and hybrid refinement. Our results demonstrate that prompt-optimized LLMs can serve as effective, scalable alternatives to traditional supervised models in biomedical NER, particularly in rare disease applications where annotated data is scarce.
Authors: Sandeep Reddy, Kabir Khan, Rohit Patil, Ananya Chakraborty, Faizan A. Khan, Swati Kulkarni, Arjun Verma, Neha Singh
Abstract: Large language models (LLMs) are limited by substantial computational cost. We introduce a "computational economics" framework that treats an LLM as an internal economy of resource-constrained agents (attention heads and neuron blocks) that must allocate scarce computation to maximize task utility. First, we show empirically that when computation is scarce, standard LLMs reallocate attention toward high-value tokens while preserving accuracy. Building on this observation, we propose an incentive-driven training paradigm that augments the task loss with a differentiable computation cost term, encouraging sparse and efficient activations. On GLUE (MNLI, STS-B, CoLA) and WikiText-103, the method yields a family of models that trace a Pareto frontier and consistently dominate post-hoc pruning; for a similar accuracy we obtain roughly a forty percent reduction in FLOPS and lower latency, together with more interpretable attention patterns. These results indicate that economic principles offer a principled route to designing efficient, adaptive, and more transparent LLMs under strict resource constraints.
Authors: Emanuel Z. Fenech-Borg, Tilen P. Meznaric-Kos, Milica D. Lekovic-Bojovic, Arni J. Hentze-Djurhuus
Abstract: Large language models (LLMs) are deployed globally, yet their underlying cultural and ethical assumptions remain underexplored. We propose the notion of a "cultural gene" -- a systematic value orientation that LLMs inherit from their training corpora -- and introduce a Cultural Probe Dataset (CPD) of 200 prompts targeting two classic cross-cultural dimensions: Individualism-Collectivism (IDV) and Power Distance (PDI). Using standardized zero-shot prompts, we compare a Western-centric model (GPT-4) and an Eastern-centric model (ERNIE Bot). Human annotation shows significant and consistent divergence across both dimensions. GPT-4 exhibits individualistic and low-power-distance tendencies (IDV score approx 1.21; PDI score approx -1.05), while ERNIE Bot shows collectivistic and higher-power-distance tendencies (IDV approx -0.89; PDI approx 0.76); differences are statistically significant (p < 0.001). We further compute a Cultural Alignment Index (CAI) against Hofstede's national scores and find GPT-4 aligns more closely with the USA (e.g., IDV CAI approx 0.91; PDI CAI approx 0.88) whereas ERNIE Bot aligns more closely with China (IDV CAI approx 0.85; PDI CAI approx 0.81). Qualitative analyses of dilemma resolution and authority-related judgments illustrate how these orientations surface in reasoning. Our results support the view that LLMs function as statistical mirrors of their cultural corpora and motivate culturally aware evaluation and deployment to avoid algorithmic cultural hegemony.
Authors: Kaviraj Pather, Elena Hadjigeorgiou, Arben Krasniqi, Claire Schmit, Irina Rusu, Marc Pons, Kabir Khan
Abstract: Large language models (LLMs) show strong reasoning via chain-of-thought (CoT) prompting, but the process is opaque, which makes verification, debugging, and control difficult in high-stakes settings. We present Vis-CoT, a human-in-the-loop framework that converts linear CoT text into an interactive reasoning graph. Users can visualize the logical flow, identify flawed steps, and intervene by pruning incorrect paths and grafting new, user-defined premises. This shifts interaction from passive observation to active collaboration, steering models toward more accurate and trustworthy conclusions. Across GSM8K and StrategyQA, Vis-CoT improves final-answer accuracy by up to 24 percentage points over non-interactive baselines. A user study also shows large gains in perceived usability and trust. Vis-CoT points to a practical path for more reliable, understandable, and collaborative reasoning by combining LLMs with targeted human oversight.
Authors: Markus Oehri, Giulia Conti, Kaviraj Pather, Alexandre Rossi, Laia Serra, Adrian Parody, Rogvi Johannesen, Aviaja Petersen, Arben Krasniqi
Abstract: Deployed language models must decide not only what to answer but also when not to answer. We present UniCR, a unified framework that turns heterogeneous uncertainty evidence including sequence likelihoods, self-consistency dispersion, retrieval compatibility, and tool or verifier feedback into a calibrated probability of correctness and then enforces a user-specified error budget via principled refusal. UniCR learns a lightweight calibration head with temperature scaling and proper scoring, supports API-only models through black-box features, and offers distribution-free guarantees using conformal risk control. For long-form generation, we align confidence with semantic fidelity by supervising on atomic factuality scores derived from retrieved evidence, reducing confident hallucinations while preserving coverage. Experiments on short-form QA, code generation with execution tests, and retrieval-augmented long-form QA show consistent improvements in calibration metrics, lower area under the risk-coverage curve, and higher coverage at fixed risk compared to entropy or logit thresholds, post-hoc calibrators, and end-to-end selective baselines. Analyses reveal that evidence contradiction, semantic dispersion, and tool inconsistency are the dominant drivers of abstention, yielding informative user-facing refusal messages. The result is a portable recipe of evidence fusion to calibrated probability to risk-controlled decision that improves trustworthiness without fine-tuning the base model and remains valid under distribution shift.
Authors: Thanh-Long V. Le, Myeongho Jeon, Kim Vu, Viet Lai, Eunho Yang
Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is a powerful framework for improving the reasoning abilities of Large Language Models (LLMs). However, current methods such as GRPO rely only on problems where the model responses to the same input differ in correctness, while ignoring those where all responses receive the same reward -- so-called zero-variance prompts. In this work, we argue that such prompts are not useless but can, in fact, provide meaningful feedback for policy optimization. To this end, we introduce RL with Zero-Variance Prompts (RL-ZVP), a novel algorithm that extract learning signals from zero-variance prompts. RL-ZVP directly rewards correctness and penalizes errors even without contrasting responses, modulating feedback with token-level characteristics to preserve informative, nuanced signals. Across six math reasoning benchmarks, RL-ZVP achieves significant improvements of up to 8.61 points in accuracy and 7.77 points in pass rate over GRPO, while consistently outperforming other baselines that filter out zero-variance prompts. These results highlight the untapped potential of learning from zero-variance prompts in RLVR.
Authors: Giovanni Monea, Yair Feldman, Shankar Padmanabhan, Kiant\'e Brantley, Yoav Artzi
Abstract: The scalability of large language models for long-context reasoning is severely constrained by the linear growth of their Transformer key-value cache, which incurs significant memory and computational costs. We posit that as a model generates reasoning tokens, the informational value of past generated tokens diminishes, creating an opportunity for compression. In this work, we propose to periodically compress the generation KV cache with a learned, special-purpose token and evict compressed entries. We train the model to perform this compression via a modified joint distillation and reinforcement learning (RL) framework. Our training method minimizes overhead over the conventional RL process, as it leverages RL outputs for distillation. Empirically, our method achieves a superior memory-accuracy Pareto frontier compared to both the model without cache compression and training-free compression techniques.
Authors: Quan Nguyen-Tri, Mukul Ranjan, Zhiqiang Shen
Abstract: This work studies how to adaptively recompute key-value (KV) caches for diffusion large language models (DLMs) to maximize prediction accuracy while minimizing decoding latency. Prior methods' decoders recompute QKV for all tokens at every denoising step and layer, despite KV states changing little across most steps, especially in shallow layers, leading to substantial redundancy. We make three observations: (1) distant ${\bf MASK}$ tokens primarily act as a length-bias and can be cached block-wise beyond the active prediction window; (2) KV dynamics increase with depth, suggesting that selective refresh starting from deeper layers is sufficient; and (3) the most-attended token exhibits the smallest KV drift, providing a conservative lower bound on cache change for other tokens. Building on these, we propose ${\bf Elastic-Cache}$, a training-free, architecture-agnostic strategy that jointly decides ${when}$ to refresh (via an attention-aware drift test on the most-attended token) and ${where}$ to refresh (via a depth-aware schedule that recomputes from a chosen layer onward while reusing shallow-layer caches and off-window MASK caches). Unlike fixed-period schemes, Elastic-Cache performs adaptive, layer-aware cache updates for diffusion LLMs, reducing redundant computation and accelerating decoding with negligible loss in generation quality. Experiments on LLaDA-Instruct, LLaDA-1.5, and LLaDA-V across mathematical reasoning and code generation tasks demonstrate consistent speedups: $8.7\times$ on GSM8K (256 tokens), and $45.1\times$ on longer sequences, while consistently maintaining higher accuracy than the baseline. Our method achieves significantly higher throughput ($6.8\times$ on GSM8K) than existing confidence-based approaches while preserving generation quality, enabling practical deployment of diffusion LLMs.
Authors: Sibo Xiao, Jinyuan Fu, Zhongle Xie, Lidan Shou
Abstract: Accelerating the inference of large language models (LLMs) has been a critical challenge in generative AI. Speculative decoding (SD) substantially improves LLM inference efficiency. However, its utility is limited by a fundamental constraint: the draft and target models must share the same vocabulary, thus limiting the herd of available draft models and often necessitating the training of a new model from scratch. Inspired by Dynamic Time Warping (DTW), a classic algorithm for aligning time series, we propose the algorithm TokenTiming for universal speculative decoding. It operates by re-encoding the draft token sequence to get a new target token sequence, and then uses DTW to build a mapping to transfer the probability distributions for speculative sampling. Benefiting from this, our method accommodates mismatched vocabularies and works with any off-the-shelf models without retraining and modification. We conduct comprehensive experiments on various tasks, demonstrating 1.57x speedup. This work enables a universal approach for draft model selection, making SD a more versatile and practical tool for LLM acceleration.
Authors: Jinliang Liu, Jiale Bai, Shaoning Zeng
Abstract: Large language models (LLMs) excel at language understanding but often hallucinate and struggle with multi-hop reasoning. Knowledge-graph-based retrieval-augmented generation (KG-RAG) offers grounding, yet most methods rely on flat embeddings and noisy path exploration. We propose ParallaxRAG, a framework that symmetrically decouples queries and graph triples into multi-view spaces, enabling a robust retrieval architecture that explicitly enforces head diversity while constraining weakly related paths. Central to our approach is the observation that different attention heads specialize in semantic relations at distinct reasoning stages, contributing to different hops of the reasoning chain. This specialization allows ParallaxRAG to construct cleaner subgraphs and guide LLMs through grounded, step-wise reasoning. Experiments on WebQSP and CWQ, under our unified, reproducible setup (BGE-M3 + Llama3.1-8B), demonstrate competitive retrieval and QA performance, alongside reduced hallucination and good generalization. Our results highlight multi-view head specialization as a principled direction for knowledge-grounded multi-hop reasoning. Our implementation will be released as soon as the paper is accepted.
Authors: Qiang Ding, Lvzhou Luo, Yixuan Cao, Ping Luo
Abstract: Ensuring that Large Language Models (LLMs) generate summaries faithful to a given source document is essential for real-world applications. While prior research has explored LLM faithfulness, existing benchmarks suffer from annotation ambiguity, primarily due to the ill-defined boundary of permissible external knowledge in generated outputs. For instance, common sense is often incorporated into responses and labeled as "faithful", yet the acceptable extent of such knowledge remains unspecified, leading to inconsistent annotations. To address this issue, we propose a novel faithfulness annotation framework, which introduces an intermediate category, Out-Dependent, to classify cases where external knowledge is required for verification. Using this framework, we construct VeriGray (Verification with the Gray Zone) -- a new unfaithfulness detection benchmark in summarization. Statistics reveal that even SOTA LLMs, such as GPT-5, exhibit hallucinations ($\sim 6\%$ of sentences) in summarization tasks. Moreover, a substantial proportion ($\sim 9\%$ on average of models) of generated sentences fall into the Out-Dependent category, underscoring the importance of resolving annotation ambiguity in unfaithfulness detection benchmarks. Experiments demonstrate that our benchmark poses significant challenges to multiple baseline methods, indicating considerable room for future improvement.
Authors: Haoyuan Li, Yuanbo Tong, Yuchen Li, Zirui Wang, Chunhou Liu, Jiamou Liu
Abstract: Personality recognition from text is typically cast as hard-label classification, which obscures the graded, prototype-like nature of human personality judgments. We present ProtoMBTI, a cognitively aligned framework for MBTI inference that operationalizes prototype theory within an LLM-based pipeline. First, we construct a balanced, quality-controlled corpus via LLM-guided multi-dimensional augmentation (semantic, linguistic, sentiment). Next, we LoRA-fine-tune a lightweight (<=2B) encoder to learn discriminative embeddings and to standardize a bank of personality prototypes. At inference, we retrieve top-k prototypes for a query post and perform a retrieve--reuse--revise--retain cycle: the model aggregates prototype evidence via prompt-based voting, revises when inconsistencies arise, and, upon correct prediction, retains the sample to continually enrich the prototype library. Across Kaggle and Pandora benchmarks, ProtoMBTI improves over baselines on both the four MBTI dichotomies and the full 16-type task, and exhibits robust cross-dataset generalization. Our results indicate that aligning the inference process with psychological prototype reasoning yields gains in accuracy, interpretability, and transfer for text-based personality modeling.
Authors: Wenjin Liu, Haoran Luo, Xueyuan Lin, Haoming Liu, Tiesunlong Shen, Jiapu Wang, Rui Mao, Erik Cambria
Abstract: Recently, advanced large language models (LLMs) have emerged at an increasingly rapid pace. However, when faced with complex problems, most users are often unable to provide accurate and effective prompts to interact with LLMs, thus limiting the performance of LLMs. To address this challenge, we propose Prompt-R1, an end-to-end reinforcement learning framework that uses a small-scale LLM to collaborate with large-scale LLMs, replacing user interaction to solve problems better. This collaboration is cast as a multi-turn prompt interaction, where the small-scale LLM thinks and generates prompts, and the large-scale LLM performs complex reasoning. A dual-constrained reward is designed to optimize for correctness, generation quality, and reasoning accuracy. Prompt-R1 provides a plug-and-play framework that supports both inference and training with various large-scale LLMs. Experiments on multiple public datasets show that Prompt-R1 significantly outperforms baseline models across tasks. Our code is publicly available at https://github.com/QwenQKing/Prompt-R1.
Authors: Kaiyuan Zhang, Chenghao Yang, Zhoufutu Wen, Sihang Yuan, Qiuyue Wang, Chaoyi Huang, Guosheng Zhu, He Wang, Huawenyu Lu, Jianing Wen, Jianpeng Jiao, Lishu Luo, Longxiang Liu, Sijin Wu, Xiaolei Zhu, Xuanliang Zhang, Yu Liu, Ge Zhang, Yi Lin, Guang Shi, Chaoyou Fu, Wenhao Huang
Abstract: As reasoning models scale rapidly, the essential role of multimodality in human cognition has come into sharp relief, driving a growing need to probe vision-centric cognitive behaviors. Yet, existing multimodal benchmarks either overemphasize textual reasoning or fall short of systematically capturing vision-centric cognitive behaviors, leaving the cognitive capacity of MLLMs insufficiently assessed. To address this limitation, we introduce MME-CC (Multi-Modal Evaluation benchmark of Cognitive Capacity), a vision-grounded benchmark that organizes 11 representative reasoning tasks into three fundamental categories of visual information: spatial, geometric, and knowledge-based reasoning, and provides fine-grained analyses of MLLMs' cognitive capacity across these dimensions. Based on MME-CC, we conduct extensive experiments over 16 representative MLLMs. Our study reveals that closed-source models currently lead overall (e.g., 42.66 for Gemini-2.5-Pro vs. 30.45 for GLM-4.5V), while spatial and geometric reasoning remain broadly weak (less than or equal to 30%). We further identify common error patterns, including orientation mistakes, fragile cross-view identity persistence, and poor adherence to counterfactual instructions, and observe that Chain-of-Thought typically follows a three-stage process (extract -> reason -> verify) with heavy reliance on visual extraction. We hope this work catalyzes a shift toward treating the cognitive capacity of MLLMs as central to both evaluation and model design.
Authors: Steven Wang, Kyle Hunt, Shaojie Tang, Kenneth Joseph
Abstract: There is ongoing debate about whether large language models (LLMs) can serve as substitutes for human participants in survey and experimental research. While recent work in fields such as marketing and psychology has explored the potential of LLM-based simulation, a growing body of evidence cautions against this practice: LLMs often fail to align with real human behavior, exhibiting limited diversity, systematic misalignment for minority subgroups, insufficient within-group variance, and discrepancies between stated beliefs and actions. This study examines an important and distinct question in this domain: whether fine-tuning on a small subset of human survey data, such as that obtainable from a pilot study, can mitigate these issues and yield realistic simulated outcomes. Using a behavioral experiment on information disclosure, we compare human and LLM-generated responses across multiple dimensions, including distributional divergence, subgroup alignment, belief-action coherence, and the recovery of regression coefficients. We find that fine-tuning on small human samples substantially improves heterogeneity, alignment, and belief-action coherence relative to the base model. However, even the best-performing fine-tuned models fail to reproduce the regression coefficients of the original study, suggesting that LLM-generated data remain unsuitable for replacing human participants in formal inferential analyses.
Authors: Yixing Xu, Chao Li, Xuanwu Yin, Spandan Tiwari, Dong Li, Ashish Sirasao, Emad Barsoum
Abstract: Low-rank adaptation (LoRA) is one of the most popular methods among parameter-efficient fine-tuning (PEFT) methods to adapt pre-trained large language models (LLMs) to specific downstream tasks. However, the model trained based on LoRA often has an unsatisfactory performance due to its low-rank assumption. In this paper, we propose a novel method called Dual LoRA to improve the performance by incorporating an inductive bias into the original LoRA. Specifically, we separate low-rank matrices into two groups: the magnitude group to control whether or not and how far we should update a parameter and the direction group to decide whether this parameter should move forward or backward, to better simulate the parameter updating process of the full fine-tuning based on gradient-based optimization algorithms. We show that this can be simply achieved by adding a ReLU function to the magnitude group and a sign function to the direction group. We conduct several experiments over a wide range of NLP tasks, including natural language understanding (NLU) and commonsense reasoning datasets on RoBERTa, DeBERTa, and LLaMA-1/2/3 as baseline models. The results show that we consistently outperform LoRA and its state-of-the-art variants with the same number of trainable parameters.
Authors: Jihyung Park, Saleh Afroogh, Junfeng Jiao
Abstract: Large Language Models (LLM) are increasingly integrated into everyday interactions, serving not only as information assistants but also as emotional companions. Even in the absence of explicit toxicity, repeated emotional reinforcement or affective drift can gradually escalate distress in a form of \textit{implicit harm} that traditional toxicity filters fail to detect. Existing guardrail mechanisms often rely on external classifiers or clinical rubrics that may lag behind the nuanced, real-time dynamics of a developing conversation. To address this gap, we propose GAUGE (Guarding Affective Utterance Generation Escalation), logit-based framework for the real-time detection of hidden conversational escalation. GAUGE measures how an LLM's output probabilistically shifts the affective state of a dialogue.
Authors: Navid Asgari, Benjamin M. Cole
Abstract: In this study, we propose a structured methodology that utilizes large language models (LLMs) in a cost-efficient and parsimonious manner, integrating the strengths of scholars and machines while offsetting their respective weaknesses. Our methodology, facilitated through a chain of thought and few-shot learning prompting from computer science, extends best practices for co-author teams in qualitative research to human-machine teams in quantitative research. This allows humans to utilize abductive reasoning and natural language to interrogate not just what the machine has done but also what the human has done. Our method highlights how scholars can manage inherent weaknesses OF LLMs using careful, low-cost techniques. We demonstrate how to use the methodology to interrogate human-machine rating discrepancies for a sample of 1,934 press releases announcing pharmaceutical alliances (1990-2017).
Authors: Aheli Poddar (Institute of Engineering & Management, Kolkata), Saptarshi Sahoo (Indian Statistical Institute, Chennai), Sujata Ghosh (Indian Statistical Institute, Chennai)
Abstract: We study syllogistic reasoning in LLMs from the logical and natural language perspectives. In process, we explore fundamental reasoning capabilities of the LLMs and the direction this research is moving forward. To aid in our studies, we use 14 large language models and investigate their syllogistic reasoning capabilities in terms of symbolic inferences as well as natural language understanding. Even though this reasoning mechanism is not a uniform emergent property across LLMs, the perfect symbolic performances in certain models make us wonder whether LLMs are becoming more and more formal reasoning mechanisms, rather than making explicit the nuances of human reasoning.
Authors: Marcus Ma, Cole Johnson, Nolan Bridges, Jackson Trager, Georgios Chochlakis, Shrikanth Narayanan
Abstract: Third-party annotation is the status quo for labeling text, but egocentric information such as sentiment and belief can at best only be approximated by a third-person proxy. We introduce author labeling, an annotation technique where the writer of the document itself annotates the data at the moment of creation. We collaborate with a commercial chatbot with over 20,000 users to deploy an author labeling annotation system. This system identifies task-relevant queries, generates on-the-fly labeling questions, and records authors' answers in real time. We train and deploy an online-learning model architecture for product recommendation with author-labeled data to improve performance. We train our model to minimize the prediction error on questions generated for a set of predetermined subjective beliefs using author-labeled responses. Our model achieves a 537% improvement in click-through rate compared to an industry advertising baseline running concurrently. We then compare the quality and practicality of author labeling to three traditional annotation approaches for sentiment analysis and find author labeling to be higher quality, faster to acquire, and cheaper. These findings reinforce existing literature that annotations, especially for egocentric and subjective beliefs, are significantly higher quality when labeled by the author rather than a third party. To facilitate broader scientific adoption, we release an author labeling service for the research community at https://academic.echollm.io.
Authors: Yiqing Zhou, Yu Lei, Shuzheng Si, Qingyan Sun, Wei Wang, Yifei Wu, Hao Wen, Gang Chen, Fanchao Qi, Maosong Sun
Abstract: Managing extensive context remains a critical bottleneck for Large Language Models (LLMs), particularly in applications like long-document question answering and autonomous agents where lengthy inputs incur high computational costs and introduce noise. Existing compression techniques often disrupt local coherence through discrete token removal or rely on implicit latent encoding that suffers from positional bias and incompatibility with closed-source APIs. To address these limitations, we introduce the EDU-based Context Compressor, a novel explicit compression framework designed to preserve both global structure and fine-grained details. Our approach reformulates context compression as a structure-then-select process. First, our LingoEDU transforms linear text into a structural relation tree of Elementary Discourse Units (EDUs) which are anchored strictly to source indices to eliminate hallucination. Second, a lightweight ranking module selects query-relevant sub-trees for linearization. To rigorously evaluate structural understanding, we release StructBench, a manually annotated dataset of 248 diverse documents. Empirical results demonstrate that our method achieves state-of-the-art structural prediction accuracy and significantly outperforms frontier LLMs while reducing costs. Furthermore, our structure-aware compression substantially enhances performance across downstream tasks ranging from long-context tasks to complex Deep Search scenarios.
Authors: Yuki Nakayama, Koki Hikichi, Yun Ching Liu, Yu Hirate
Abstract: This paper presents a large-scale corpus of Rakuten Travel Reviews. Our collection contains 7.29 million customer reviews for 16 years, ranging from 2009 to 2024. Each record in the dataset contains the review text, its response from an accommodation, an anonymized reviewer ID, review date, accommodation ID, plan ID, plan title, room type, room name, purpose, accompanying group, and user ratings from six aspect categories, as well as an overall score. We present statistical information about our corpus and provide insights into factors driving data drift between 2019 and 2024 using statistical approaches.
Authors: Tung Duong Ta, Tim Oates, Thien Van Luong, Huan Vu, Tien Cuong Nguyen
Abstract: Despite advances in mathematical reasoning capabilities, Large Language Models (LLMs) still struggle with calculation verification when using established prompting techniques. We present MDToC (Metacognitive Dynamic Tree of Concepts), a three-phase approach that constructs a concept tree, develops accuracy-verified calculations for each concept, and employs majority voting to evaluate competing solutions. Evaluations across CHAMP, MATH, and Game-of-24 benchmarks demonstrate our MDToC's effectiveness, with GPT-4-Turbo achieving 58.1\% on CHAMP, 86.6\% on MATH, and 85\% on Game-of-24 - outperforming GoT by 5\%, 5.4\%, and 4\% on all these tasks, respectively, without hand-engineered hints. MDToC consistently surpasses existing prompting methods across all backbone models, yielding improvements of up to 7.6\% over ToT and 6.2\% over GoT, establishing metacognitive calculation verification as a promising direction for enhanced mathematical reasoning.
Authors: Tongyi Fun Team, Qian Chen, Luyao Cheng, Chong Deng, Xiangang Li, Jiaqing Liu, Chao-Hong Tan, Wen Wang, Junhao Xu, Jieping Ye, Qinglin Zhang, Qiquan Zhang, Jingren Zhou
Abstract: Recent advancements in joint speech-text models show great potential for seamless voice interactions. However, existing models face critical challenges: temporal resolution mismatch between speech tokens (25Hz) and text tokens (~3Hz) dilutes semantic information, incurs high computational costs, and causes catastrophic forgetting of text LLM knowledge. We introduce Fun-Audio-Chat, a Large Audio Language Model addressing these limitations via two innovations from our previous work DrVoice. First, Dual-Resolution Speech Representations (DRSR): the Shared LLM processes audio at efficient 5Hz (via token grouping), while the Speech Refined Head generates high-quality tokens at 25Hz, balancing efficiency (~50% GPU reduction) and quality. Second, Core-Cocktail Training, a two-stage fine-tuning with intermediate merging that mitigates catastrophic forgetting. We then apply Multi-Task DPO Training to enhance robustness, audio understanding, instruction-following and voice empathy. This multi-stage post-training enables Fun-Audio-Chat to retain text LLM knowledge while gaining powerful audio understanding, reasoning, and generation. Unlike recent LALMs requiring large-scale audio-text pre-training, Fun-Audio-Chat leverages pre-trained models and extensive post-training. Fun-Audio-Chat 8B and MoE 30B-A3B achieve competitive performance on Speech-to-Text and Speech-to-Speech tasks, ranking top among similar-scale models on Spoken QA benchmarks. They also achieve competitive to superior performance on Audio Understanding, Speech Function Calling, Instruction-Following and Voice Empathy. We develop Fun-Audio-Chat-Duplex, a full-duplex variant with strong performance on Spoken QA and full-duplex interactions. We open-source Fun-Audio-Chat-8B with training and inference code, and provide an interactive demo, at https://github.com/FunAudioLLM/Fun-Audio-Chat .
Authors: Chen Hu, Haikuo Du, Heng Wang, Lin Lin, Mingrui Chen, Peng Liu, Ruihang Miao, Tianchi Yue, Wang You, Wei Ji, Wei Yuan, Wenjin Deng, Xiaojian Yuan, Xiaoyun Zhang, Xiangyu Liu, Xikai Liu, Yanming Xu, Yicheng Cao, Yifei Zhang, Yongyao Wang, Yubo Shu, Yurong Zhang, Yuxiang Zhang, Zheng Gong, Zhichao Chang, Binyan Li, Dan Ma, Furong Jia, Hongyuan Wang, Jiayu Liu, Jing Bai, Junlan Liu, Manjiao Liu, Na Wang, Qiuping Wu, Qinxin Du, Shiwei Li, Wen Sun, Yifeng Gong, Yonglin Chen, Yuling Zhao, Yuxuan Lin, Ziqi Ren, Zixuan Wang, Aihu Zhang, Brian Li, Buyun Ma, Kang An, Li Xie, Mingliang Li, Pan Li, Shidong Yang, Xi Chen, Xiaojia Liu, Yuchu Luo, Yuan Song, YuanHao Ding, Yuanwei Liang, Zexi Li, Zhaoning Zhang, Zixin Zhang, Binxing Jiao, Daxin Jiang, Jiansheng Chen, Jing Li, Xiangyu Zhang, Yibo Zhu
Abstract: As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.
Authors: Alexander Podolskiy, Semen Molokov, Timofey Gerasin, Maksim Titov, Alexey Rukhovich, Artem Khrapov, Kirill Morozov, Evgeny Tetin, Constantine Korikov, Pavel Efimov, Polina Lazukova, Yuliya Skripkar, Nikita Okhotnikov, Irina Piontkovskaya, Meng Xiaojun, Zou Xueyi, Zhang Zhenhe
Abstract: We present Gamayun, a 1.5B-parameter multilingual language model trained entirely from scratch on 2.5T tokens. Designed for efficiency and deployment in resource-constrained environments, Gamayun addresses the lack of research on small non-English-centric LLMs by adopting a novel two-stage pre-training strategy: balanced multilingual training for cross-lingual alignment, followed by high-quality English enrichment to transfer performance gains across languages. Our model supports 12 languages, with special focus on Russian. Despite a significantly smaller training budget than comparable models, Gamayun outperforms LLaMA3.2-1B (9T tokens) on all considered benchmarks, and surpasses Qwen2.5-1.5B (18T tokens) on a wide range of English and multilingual tasks. It matches or exceeds Qwen3 (36T tokens) on most tasks outside advanced STEM, achieving state-of-the-art results in Russian, including the MERA benchmark, among the models of comparable size (1-2B parameters).
Authors: Yi Liu, Gelei Deng, Yuekang Li, Kailong Wang, Zihao Wang, Xiaofeng Wang, Tianwei Zhang, Yepang Liu, Haoyu Wang, Yan Zheng, Leo Yu Zhang, Yang Liu
Abstract: Large Language Models (LLMs), renowned for their superior proficiency in language comprehension and generation, stimulate a vibrant ecosystem of applications around them. However, their extensive assimilation into various services introduces significant security risks. This study deconstructs the complexities and implications of prompt injection attacks on actual LLM-integrated applications. Initially, we conduct an exploratory analysis on ten commercial applications, highlighting the constraints of current attack strategies in practice. Prompted by these limitations, we subsequently formulate HouYi, a novel black-box prompt injection attack technique, which draws inspiration from traditional web injection attacks. HouYi is compartmentalized into three crucial elements: a seamlessly-incorporated pre-constructed prompt, an injection prompt inducing context partition, and a malicious payload designed to fulfill the attack objectives. Leveraging HouYi, we unveil previously unknown and severe attack outcomes, such as unrestricted arbitrary LLM usage and uncomplicated application prompt theft. We deploy HouYi on 36 actual LLM-integrated applications and discern 31 applications susceptible to prompt injection. 10 vendors have validated our discoveries, including Notion, which has the potential to impact millions of users. Our investigation illuminates both the possible risks of prompt injection attacks and the possible tactics for mitigation.
Authors: Jianing He, Qi Zhang, Weiping Ding, Duoqian Miao, Jun Zhao, Liang Hu, Longbing Cao
Abstract: Early exiting has demonstrated its effectiveness in accelerating the inference of pre-trained language models like BERT by dynamically adjusting the number of layers executed. However, most existing early exiting methods only consider local information from an individual test sample to determine their exiting indicators, failing to leverage the global information offered by sample population. This leads to suboptimal estimation of prediction correctness, resulting in erroneous exiting decisions. To bridge the gap, we explore the necessity of effectively combining both local and global information to ensure reliable early exiting during inference. Purposefully, we leverage prototypical networks to learn class prototypes and devise a distance metric between samples and class prototypes. This enables us to utilize global information for estimating the correctness of early predictions. On this basis, we propose a novel Distance-Enhanced Early Exiting framework for BERT (DE$^3$-BERT). DE$^3$-BERT implements a hybrid exiting strategy that supplements classic entropy-based local information with distance-based global information to enhance the estimation of prediction correctness for more reliable early exiting decisions. Extensive experiments on the GLUE benchmark demonstrate that DE$^3$-BERT consistently outperforms state-of-the-art models under different speed-up ratios with minimal storage or computational overhead, yielding a better trade-off between model performance and inference efficiency. Additionally, an in-depth analysis further validates the generality and interpretability of our method.
Authors: Kavana Venkatesh, Yusuf Dalva, Ismini Lourentzou, Pinar Yanardag
Abstract: Despite impressive visual fidelity, current text-to-image (T2I) diffusion models struggle to depict rare, complex, or culturally nuanced concepts due to training data limitations. We introduce RAVEL, a training-free framework that significantly improves rare concept generation, context-driven image editing, and self-correction by integrating graph-based retrieval-augmented generation (RAG) into diffusion pipelines. Unlike prior RAG and LLM-enhanced methods reliant on visual exemplars, static captions or pre-trained knowledge of models, RAVEL leverages structured knowledge graphs to retrieve compositional, symbolic, and relational context, enabling nuanced grounding even in the absence of visual priors. To further refine generation quality, we propose SRD, a novel self-correction module that iteratively updates prompts via multi-aspect alignment feedback, enhancing attribute accuracy, narrative coherence, and semantic fidelity. Our framework is model-agnostic and compatible with leading diffusion models including Stable Diffusion XL, Flux, and DALL-E 3. We conduct extensive evaluations across three newly proposed benchmarks - MythoBench, Rare-Concept-1K, and NovelBench. RAVEL also consistently outperforms SOTA methods across perceptual, alignment, and LLM-as-a-Judge metrics. These results position RAVEL as a robust paradigm for controllable and interpretable T2I generation in long-tail domains.
Authors: Sicheng Zhu, Brandon Amos, Yuandong Tian, Chuan Guo, Ivan Evtimov
Abstract: Many jailbreak attacks on large language models (LLMs) rely on a common objective: making the model respond with the prefix ``Sure, here is (harmful request)''. While straightforward, this objective has two limitations: limited control over model behaviors, yielding incomplete or unrealistic jailbroken responses, and a rigid format that hinders optimization. We introduce AdvPrefix, a plug-and-play prefix-forcing objective that selects one or more model-dependent prefixes by combining two criteria: high prefilling attack success rates and low negative log-likelihood. AdvPrefix integrates seamlessly into existing jailbreak attacks to mitigate the previous limitations for free. For example, replacing GCG's default prefixes on Llama-3 improves nuanced attack success rates from 14% to 80%, revealing that current safety alignment fails to generalize to new prefixes. Code and selected prefixes are released at github.com/facebookresearch/jailbreak-objectives.
Authors: Xindi Wu, Mengzhou Xia, Rulin Shao, Zhiwei Deng, Pang Wei Koh, Olga Russakovsky
Abstract: Training vision-language models via instruction tuning relies on large data mixtures spanning diverse tasks and domains, yet these mixtures frequently include redundant information that increases computational costs without proportional gains. Existing methods typically rely on task-agnostic heuristics to estimate data importance, limiting their effectiveness across tasks. We introduce ICONS, a gradient-based Influence CONsensus approach for vision-language data Selection. Our method leverages first-order training dynamics to estimate each example's influence on validation performance, then aggregates these estimates across tasks via majority voting. This cross-task consensus identifies consistently valuable data points while mitigating score calibration and outlier sensitivity, enabling robust and scalable data selection for diverse multitask mixtures. Models trained on our selected 20% data subset from LLAVA-665K (respectively: from CAMBRIAN-7M, from VISION-FLAN-186K) retain 98.6% (respectively: 98.8%, 99.8%) of full-dataset performance. We demonstrate that our selected data generalizes to unseen tasks and model architectures, and release three compact subsets LLAVA-ICONS-133K, CAMBRIAN-ICONS-1.4M, and VISION-FLAN-ICONS-37K for efficient vision-language model development.
Authors: Christopher Burger
Abstract: Adversarial attacks challenge the reliability of Explainable AI (XAI) by altering explanations while the model's output remains unchanged. The success of these attacks on text-based XAI is often judged using standard information retrieval metrics. We argue these measures are poorly suited in the evaluation of trustworthiness, as they treat all word perturbations equally while ignoring synonymity, which can misrepresent an attack's true impact. To address this, we apply synonymity weighting, a method that amends these measures by incorporating the semantic similarity of perturbed words. This produces more accurate vulnerability assessments and provides an important tool for assessing the robustness of AI systems. Our approach prevents the overestimation of attack success, leading to a more faithful understanding of an XAI system's true resilience against adversarial manipulation.
Authors: Terrance Yu-Hao Chen, Yulin Chen, Pontus Soederhaell, Sadrishya Agrawal, Kateryna Shapovalenko
Abstract: Decoding speech from non-invasive brain signals, such as electroencephalography (EEG), has the potential to advance brain-computer interfaces (BCIs), with applications in silent communication and assistive technologies for individuals with speech impairments. However, EEG-based speech decoding faces major challenges, such as noisy data, limited datasets, and poor performance on complex tasks like speech perception. This study attempts to address these challenges by employing variational autoencoders (VAEs) for EEG data augmentation to improve data quality and applying a state-of-the-art (SOTA) sequence-to-sequence deep learning architecture, originally successful in electromyography (EMG) tasks, to EEG-based speech decoding. Additionally, we adapt this architecture for word classification tasks. Using the Brennan dataset, which contains EEG recordings of subjects listening to narrated speech, we preprocess the data and evaluate both classification and sequence-to-sequence models for EEG-to-words/sentences tasks. Our experiments show that VAEs have the potential to reconstruct artificial EEG data for augmentation. Meanwhile, our sequence-to-sequence model achieves more promising performance in generating sentences compared to our classification model, though both remain challenging tasks. These findings lay the groundwork for future research on EEG speech perception decoding, with possible extensions to speech production tasks such as silent or imagined speech.
Authors: Zhixian Zhao, Xinfa Zhu, Xinsheng Wang, Shuiyuan Wang, Xuelong Geng, Wenjie Tian, Lei Xie
Abstract: Large-scale audio language models (ALMs), such as Qwen2-Audio, are capable of comprehending diverse audio signal, performing audio analysis and generating textual responses. However, in speech emotion recognition (SER), ALMs often suffer from hallucinations, resulting in misclassifications or irrelevant outputs. To address these challenges, we propose C$^2$SER, a novel ALM designed to enhance the stability and accuracy of SER through Contextual perception and Chain of Thought (CoT). C$^2$SER integrates the Whisper encoder for semantic perception and Emotion2Vec-S for acoustic perception, where Emotion2Vec-S extends Emotion2Vec with semi-supervised learning to enhance emotional discrimination. Additionally, C$^2$SER employs a CoT approach, processing SER in a step-by-step manner while leveraging speech content and speaking styles to improve recognition. To further enhance stability, C$^2$SER introduces self-distillation from explicit CoT to implicit CoT, mitigating error accumulation and boosting recognition accuracy. Extensive experiments show that C$^2$SER outperforms existing popular ALMs, such as Qwen2-Audio and SECap, delivering more stable and precise emotion recognition. We release the training code, checkpoints, and test sets to facilitate further research.
Authors: Aarush Sinha
Abstract: Training effective dense retrieval models typically relies on hard negative (HN) examples mined from large document corpora using methods such as BM25 or cross-encoders, which require full corpus access and expensive index construction. We propose generating synthetic hard negatives directly from a provided query and positive passage, using Large Language Models(LLMs). We fine-tune DistilBERT using synthetic negatives generated by four state-of-the-art LLMs ranging from 4B to 30B parameters (Qwen3, LLaMA3, Phi4) and evaluate performance across 10 BEIR benchmark datasets. Contrary to the prevailing assumption that stronger generative models yield better synthetic data, find that our generative pipeline consistently underperforms traditional corpus-based mining strategies (BM25 and Cross-Encoder). Furthermore, we observe that scaling the generator model does not monotonically improve retrieval performance and find that the 14B parameter model outperforms the 30B model and in some settings it is the worst performing.
Authors: Cainan Davidson, Deva Ramanan, Neehar Peri
Abstract: Autonomous Vehicles (AVs) collect and pseudo-label terabytes of multi-modal data localized to HD maps during normal fleet testing. However, identifying interesting and safety-critical scenarios from uncurated driving logs remains a significant challenge. Traditional scenario mining techniques are error-prone and prohibitively time-consuming, often relying on hand-crafted structured queries. In this work, we revisit spatio-temporal scenario mining through the lens of recent vision-language models (VLMs) to detect whether a described scenario occurs in a driving log and, if so, precisely localize it in both time and space. To address this problem, we introduce RefAV, a large-scale dataset of 10,000 diverse natural language queries that describe complex multi-agent interactions relevant to motion planning derived from 1000 driving logs in the Argoverse 2 Sensor dataset. We evaluate several referential multi-object trackers and present an empirical analysis of our baselines. Notably, we find that naively repurposing off-the-shelf VLMs yields poor performance, suggesting that scenario mining presents unique challenges. Lastly, we discuss our recently held competition and share insights from the community. Our code and dataset are available at https://github.com/CainanD/RefAV/ and https://argoverse.github.io/user-guide/tasks/scenario_mining.html
URLs: https://github.com/CainanD/RefAV/, https://argoverse.github.io/user-guide/tasks/scenario_mining.html
Authors: Dong Xue, Ziyao Shao, Zhaoyang Duan, Fangzhou Liu, Bing Li, Zhongheng Zhang
Abstract: Large multimodal models (LMMs) have demonstrated significant potential in providing innovative solutions for various biomedical tasks, including pathology analysis, radiology report generation, and biomedical assistance. However, the existing multimodal biomedical AI is typically based on foundation LLMs, thus hindering the understanding of intricate medical concepts with limited medical training data. Moreover, recent LLaVA-induced medical LMMs struggle to effectively capture the intricate relationship between the texts and the images. Therefore, we introduce Doctor Sun, a large multimodal generative model specialized in medicine, developed to encode, integrate, and interpret diverse biomedical data modalities such as text and images. In particular, Doctor Sun integrates a pre-trained vision encoder with a medical LLM and conducts two-stage training on various medical datasets, focusing on feature alignment and instruction tuning. Moreover, we release SunMed-VL, a wide-range bilingual medical multimodal dataset, along with all associated models, code, and resources, to freely support the advancement of biomedical multimodal research.
Authors: Xiang Zhang, Kun Wei, Xu Yang, Chenghao Xu, Su Yan, Cheng Deng
Abstract: As Large Language Models (LLMs) become increasingly prevalent, their security vulnerabilities have already drawn attention. Machine unlearning is introduced to seek to mitigate these risks by removing the influence of undesirable data. However, existing methods not only rely on the retained dataset to preserve model utility, but also suffer from cumulative catastrophic utility loss under continuous unlearning requests. To solve this dilemma, we propose a novel method, called Rotation Control Unlearning (RCU), which leverages the rotational salience weight of RCU to quantify and control the unlearning degree in the continuous unlearning process. The skew symmetric loss is designed to construct the existence of the cognitive rotation space, where the changes of rotational angle can simulate the continuous unlearning process. Furthermore, we design an orthogonal rotation axes regularization to enforce mutually perpendicular rotation directions for continuous unlearning requests, effectively minimizing interference and addressing cumulative catastrophic utility loss. Experiments on multiple datasets confirm that our method without retained dataset achieves SOTA performance.
Authors: Hasan Akgul, Daniel Borg, Arta Berisha, Amina Rahimova, Andrej Novak, Mila Petrov
Abstract: Large language models are often adapted through parameter efficient fine tuning, but current release practices provide weak assurances about what data were used and how updates were computed. We present Verifiable Fine Tuning, a protocol and system that produces succinct zero knowledge proofs that a released model was obtained from a public initialization under a declared training program and an auditable dataset commitment. The approach combines five elements. First, commitments that bind data sources, preprocessing, licenses, and per epoch quota counters to a manifest. Second, a verifiable sampler that supports public replayable and private index hiding batch selection. Third, update circuits restricted to parameter efficient fine tuning that enforce AdamW style optimizer semantics and proof friendly approximations with explicit error budgets. Fourth, recursive aggregation that folds per step proofs into per epoch and end to end certificates with millisecond verification. Fifth, provenance binding and optional trusted execution property cards that attest code identity and constants. On English and bilingual instruction mixtures, the method maintains utility within tight budgets while achieving practical proof performance. Policy quotas are enforced with zero violations, and private sampling windows show no measurable index leakage. Federated experiments demonstrate that the system composes with probabilistic audits and bandwidth constraints. These results indicate that end to end verifiable fine tuning is feasible today for real parameter efficient pipelines, closing a critical trust gap for regulated and decentralized deployments.
Authors: Cheng Yuan, Jiawei Shao, Chi Zhang, Xuelong Li
Abstract: Recent years have witnessed the rapid advancements of large language models (LLMs) and their expanding applications, leading to soaring demands for computational resources. The widespread adoption of test-time scaling further aggravates the tension between model capability and resource consumption, highlighting the importance of inference efficiency. However, a unified metric that accurately reflects an LLM's efficiency across different model sizes and architectures remains absent. Motivated by the correlation between compression and intelligence, we introduce information capacity, a measure of model efficiency based on text compression performance relative to computational complexity. Larger models can predict the next token more accurately, achieving greater compression gains but at higher computational costs. Empirical evaluations on mainstream open-source models show that models of varying sizes within a series exhibit consistent information capacity. This metric enables a fair efficiency comparison across model series and accurate performance prediction within a model series. A distinctive feature of information capacity is that it incorporates tokenizer efficiency, which affects both input and output token counts but is often neglected in LLM evaluations. We assess the information capacity of 52 models on 5 heterogeneous datasets and observe consistent results on the influences of tokenizer efficiency, pretraining data, and the mixture-of-experts architecture.
Authors: Samuele Dell'Erba, Andrew D. Bagdanov
Abstract: Diffusion models have established the state-of-the-art in text-to-image generation, but their performance often relies on a diffusion prior network to translate text embeddings into the visual manifold for easier decoding. These priors are computationally expensive and require extensive training on massive datasets. In this work, we challenge the necessity of a trained prior at all by employing Optimization-based Visual Inversion (OVI), a training-free and zero-shot alternative, to replace the need for a prior. OVI initializes a latent visual representation from random pseudo-tokens and iteratively optimizes it to maximize the cosine similarity with the input textual prompt embedding. We further propose two novel constraints, a Mahalanobis-based and a Nearest-Neighbor loss, to regularize the OVI optimization process toward the distribution of realistic images. Our experiments, conducted on Kandinsky 2.2, show that OVI can serve as an alternative to traditional priors. More importantly, our analysis reveals a critical flaw in current evaluation benchmarks like T2I-CompBench++, where simply using the text embedding as a prior achieves surprisingly high scores, despite lower perceptual quality. Our constrained OVI methods improve visual fidelity over this baseline, with the Nearest-Neighbor approach proving particularly effective. It achieves quantitative scores comparable to or higher than the state-of-the-art data-efficient prior, underscoring the potential of optimization-based strategies as viable, training-free alternatives to traditional priors. The code will be publicly available upon acceptance.
Authors: Haipeng Luo, Huawen Feng, Qingfeng Sun, Can Xu, Kai Zheng, Yufei Wang, Tao Yang, Han Hu, Yansong Tang, Di Wang
Abstract: Large Reasoning Models (LRMs) like o3 and DeepSeek-R1 have achieved remarkable progress in natural language reasoning with long chain-of-thought. However, they remain computationally inefficient and struggle with accuracy when solving problems requiring complex mathematical operations. In this work, we present AgentMath, an agent framework that seamlessly integrates language models' reasoning capabilities with code interpreters' computational precision to efficiently tackle complex mathematical problems. Our approach introduces three key innovations: (1) An automated method that converts natural language chain-of-thought into structured tool-augmented trajectories, generating high-quality supervised fine-tuning (SFT) data to alleviate data scarcity; (2) A novel agentic reinforcement learning (RL) paradigm that dynamically interleaves natural language generation with real-time code execution. This enables models to autonomously learn optimal tool-use strategies through multi-round interactive feedback, while fostering emergent capabilities in code refinement and error correction; (3) An efficient training system incorporating innovative techniques, including request-level asynchronous rollout scheduling, agentic partial rollout, and prefix-aware weighted load balancing, achieving 4-5x speedup and making efficient RL training feasible on ultra-long sequences with scenarios with massive tool invocation. The evaluations show that AgentMath achieves state-of-the-art performance on challenging mathematical competition benchmarks including AIME24, AIME25, and HMMT25. Specifically, AgentMath-30B-A3B attains 90.6%, 86.4%, and 73.8% accuracy respectively, achieving advanced performance. The results validate the effectiveness of our approach and pave the way for building more sophisticated and scalable mathematical reasoning agents.
Authors: Ahmed M. Hussain, Salahuddin Salahuddin, Panos Papadimitratos
Abstract: Current Large Language Models (LLMs) safety approaches focus on explicitly harmful content while overlooking a critical vulnerability: the inability to understand context and recognize user intent. This creates exploitable vulnerabilities that malicious users can systematically leverage to circumvent safety mechanisms. We empirically evaluate multiple state-of-the-art LLMs, including ChatGPT, Claude, Gemini, and DeepSeek. Our analysis demonstrates the circumvention of reliable safety mechanisms through emotional framing, progressive revelation, and academic justification techniques. Notably, reasoning-enabled configurations amplified rather than mitigated the effectiveness of exploitation, increasing factual precision while failing to interrogate the underlying intent. The exception was Claude Opus 4.1, which prioritized intent detection over information provision in some use cases. This pattern reveals that current architectural designs create systematic vulnerabilities. These limitations require paradigmatic shifts toward contextual understanding and intent recognition as core safety capabilities rather than post-hoc protective mechanisms.
Authors: Jiakai Tang, Chuan Wang, Gaoming Yang, Han Wu, Jiahao Yu, Jian Wu, Jianwu Hu, Junjun Zheng, Longbin Li, Shuwen Xiao, Xiangheng Kong, Yeqiu Yang, Yuning Jiang, Ahjol Nurlanbek, Binbin Cao, Bo Zheng, Fangmei Zhu, Gaoming Zhou, Huimin Yi, Huiping Chu, Jin Huang, Jinzhe Shan, Kenan Cui, Longbin Li, Silu Zhou, Wen Chen, Xia Ming, Xiang Gao, Xin Yao, Xingyu Wen, Yan Zhang, Yiwen Hu, Yulin Wang, Ziheng Bao, Zongyuan Wu
Abstract: Industrial recommender systems face two fundamental limitations under the log-driven paradigm: (1) knowledge poverty in ID-based item representations that causes brittle interest modeling under data sparsity, and (2) systemic blindness to beyond-log user interests that constrains model performance within platform boundaries. These limitations stem from an over-reliance on shallow interaction statistics and close-looped feedback while neglecting the rich world knowledge about product semantics and cross-domain behavioral patterns that Large Language Models have learned from vast corpora. To address these challenges, we introduce ReaSeq, a reasoning-enhanced framework that leverages world knowledge in Large Language Models to address both limitations through explicit and implicit reasoning. Specifically, ReaSeq employs explicit Chain-of-Thought reasoning via multi-agent collaboration to distill structured product knowledge into semantically enriched item representations, and latent reasoning via Diffusion Large Language Models to infer plausible beyond-log behaviors. Deployed on Taobao's ranking system serving hundreds of millions of users, ReaSeq achieves substantial gains: >6.0% in IPV and CTR, >2.9% in Orders, and >2.5% in GMV, validating the effectiveness of world-knowledge-enhanced reasoning over purely log-driven approaches.