论文检索

Rachneet Kaur, Nishan Srishankar, Zhen Zeng, Sumitra Ganesh

Recent multimodal LLMs have shown promise in chart-based visual question answering, but their performance declines sharply on unannotated charts—those requiring precise visual interpretation rather than relying on textual shortcuts. To address this, we introduce ChartAgent, a novel agentic framework that explicitly performs visual reasoning directly within the chart’s spatial domain. Unlike textual chain-of-thought reasoning, ChartAgent iteratively decomposes queries into visual subtasks and actively manipulates and interacts with chart images through specialized actions such as drawing annotations, cropping regions (e.g., segmenting pie slices, isolating bars), and localizing axes, using a library of chart-specific vision tools to fulfill each subtask. This iterative reasoning process closely mirrors human cognitive strategies for chart comprehension. ChartAgent achieves state-of-the-art accuracy on the ChartBench and ChartX benchmarks, surpassing prior methods by up to 16.07% absolute gain overall and 17.31% on unannotated, numerically intensive queries. Furthermore, our analyses show that ChartAgent is (a) effective across diverse chart types, (b) achieves the highest scores across varying visual and reasoning complexity levels, and (c) serves as a plug-and-play framework that boosts performance across diverse underlying LLMs. Our work is among the first to demonstrate visually grounded reasoning for chart understanding using tool-augmented multimodal agents.

Jinbiao Wei, Yilun Zhao, Kangqi Ni, Arman Cohan

End-to-end GUI agents for real desktop environments require large amounts of high-quality interaction data, yet collecting human demonstrations is expensive and existing synthetic pipelines often suffer from limited task diversity or noisy, goal-drifting trajectories. We present a trajectory expansion framework Anchor that bootstraps scalable desktop supervision from a small set of verified seed demonstrations. Starting from each seed, we identify branch points that correspond to meaningful state changes and propose new, state-grounded task variants conditioned on the current GUI context. An executing agent then follows the proposed instructions to generate new trajectories, while a verifier enforces task completion via state-aware checks and trajectory-level consistency. To improve supervision quality, we further apply task-conditioned step-level filtering to remove ungrounded actions and denoise post-branch segments to maintain coherent intent. Experiments on standard desktop benchmarks, OSWorld and WindowsAgentArena, show that models fine-tuned on our expanded corpus achieve consistent improvements over zero-shot agents and representative synthesis baselines, and generalize across applications and operating systems.

Suyoung Bae, CheolWon Na, Jaehoon Lee, Yumin Lee, YunSeok Choi, Jee-Hyong Lee

As Large Language Models (LLMs) have become capable of generating long and descriptive code summaries, accurate and reliable evaluation of factual consistency has become a critical challenge. However, previous evaluation methods are primarily designed for short summaries of isolated code snippets. Consequently, they struggle to provide fine-grained evaluation of multi-sentence functionalities and fail to accurately assess dependency context commonly found in real-world code summaries.To address this, we propose ReFEree, a reference-free and fine-grained method for evaluating factual consistency in real-world code summaries. We define factual inconsistency criteria specific to code summaries and evaluate them at the segment level using these criteria along with dependency information. These segment-level results are then aggregated into a fine-grained score. We construct a code summarization benchmark with human-annotated factual consistency labels. The evaluation results demonstrate that ReFEree achieves the highest correlation with human judgment among 13 baselines, improving 15-18% over the previous state-of-the-art. Our code and data are available at https://github.com/bsy99615/ReFEree.git.

Yuejiao Wang, Zihao Ji, Pengfei Cai, Xu Li, Haorui Zheng, Zewen Song, Zhongliang Liu, Chen Zhang, Pengfei Wan

Recent advances in neural song generation have enabled high-quality synthesis from lyrics and global textual prompts. However, most systems fail to model temporally varying attributes of songs, severely limiting fine-grained control over musical structure and dynamics. To address this, we propose Segtune, a Diffusion Transformer-based framework enabling structured and fine-grained controllability by allowing users or large language models (LLMs) to specify local musical descriptions aligned to song segments. These segment prompts are temporally broadcast to corresponding time windows, while global prompts ensure stylistic coherence. To support precise lyric-to-music alignment, we introduce an LLM-based duration predictor that autoregressively generates sentence-level timestamps in LyRiCs format. We further construct a large-scale data pipeline for high-quality song collection with aligned lyrics and prompts, and propose new metrics to evaluate segment alignment and vocal consistency. Experiments demonstrate that Segtune outperforms existing baselines in both musicality and controllability. Visit our demo page for codes and more generated songs.

Siheng Li, Kejiao Li, Zenan Xu, Guanhua Huang, Kun Li, Haoyuan Wu, Wujiajia, Zihao Zheng, Chenchen Zhang, Kun Shi 等

Recent progress in large language models (LLMs) is largely driven by scaling training compute through either pre-training with next-token prediction (NTP) or post-training with reinforcement learning (RL). The former contributes to learning broad knowledge and skills from general data, while struggling with data inefficiency and catastrophic forgetting in continual learning settings. The latter incentivizes reasoning capabilities with strong generalization, but is constrained by limited data availability due to its reliance on human annotation. To alleviate these issues, we propose Reinforcement Learning on Pre-Training data (RLPT), which combines the advantages of learning from general data and RL. In particular, RLPT derives reward signals directly from general text data through a next-segment reasoning objective, rewarding the policy for correctly predicting next text segments conditioned on the prefix text. Experiments across multiple benchmarks and models demonstrate the effectiveness of . For example, RLPT yields substantial improvements in continual pre-training (+4.6\%) and provides a strong foundation for post-training (+3.4\%) on Qwen3-8B-Base.

Refael Shaked Greenfeld, Reut Tsarfaty

Coreference Resolution (CR) is a fundamental NLP task critical for long-form tasks as information extraction, summarization, and many business applications. However, CR methods originally designed for English struggle with Morphologically Rich Languages (MRLs), where mention boundaries do not necessarily align with word boundaries, and a single token may consist of multiple anaphors. CR modeling and evaluation protocols standardly assume that, as in English, words and mentions mostly align. However, this assumption breaks down in MRLs, particularly in the context of LLMs’ raw-text processing and end-to-end tasks. To assess and address this challenge, we introduce KibutzR, the first comprehensive CR dataset for Modern Hebrew, an MRL rich with complex words and pronominal clitics. We deliver an annotated dataset that identifies mentions at word, sub-word and multi-word levels, and propose an evaluation protocol that directly addresses word/morpheme boundary discrepancies. Our experiments show that contemporary LLMs perform significantly worse on Hebrew than on English, and that performance degrades on raw unsegmented text. Crucially, we show an inverse performance-trend in Hebrew relative to English, where smaller encoders perform far better than contemporary decoder models, leaving ample space for investigation and improvement. We deliver a new benchmark for Hebrew coreference resolution and a segmentation-aware evaluation protocol to inform future work on other MRLs.

Fabian Retkowski, Maike Züfle, Thai Binh Nguyen, Jan Niehues, Alexander Waibel

Audio chaptering, the task of automatically segmenting long-form audio into coherent sections, is increasingly important for navigating podcasts, lectures, and videos. Despite its relevance, research remains limited and text-based, leaving key questions unresolved about leveraging audio information, handling ASR errors, and transcript-free evaluation. We address these gaps through three contributions: (1) a systematic comparison between text-based models with acoustic features, a novel audio-only architecture (AudioSeg) operating on learned audio representations, and multimodal LLMs; (2) empirical analysis of factors affecting performance, including transcript quality, acoustic features, duration, and speaker composition; and (3) formalized evaluation protocols contrasting transcript-dependent text-space protocols with transcript-invariant time-space protocols. Our experiments on YTSeg reveal that AudioSeg substantially outperforms text-based approaches, pauses provide the largest acoustic gains, and current MLLMs struggle due to context limitations and weak instruction following.

Zhuoen Chen, Dongfang Li, Meishan Zhang, Baotian Hu, Min Zhang

Large Language Models (LLMs) face severe challenges in long-context processing, including quadratic computational costs, information forgetting, and the context fragmentation inherent in Retrieval-Augmented Generation (RAG). We introduce LycheeMemory, a cognitively inspired framework that enables efficient long-context inference via chunk-wise compression and selective memory recall, rather than processing all raw tokens. LycheeMemory segments the input into chunks and encodes each into compressed KV-cache style representations using a Compressor. A Gate then dynamically selects relevant memory blocks, which a Reasoner iteratively processes with an evolving working memory to solve downstream tasks. The Compressor and Reasoner are jointly optimized via end-to-end reinforcement learning, while the Gate is trained separately as a classifier. Experimental results demonstrate that LycheeMemory achieves competitive accuracy (up to 82% in ablation variants) on multi-hop reasoning benchmarks (e.g., RULER-HQA), successfully extrapolates context length from 7K to 1.75M, and provides a favorable accuracy–efficiency trade-off against strong long-context baselines. Notably, compared to MemAgent, LycheeMemory achieves an average 2× reduction in peak GPU memory usage and a 6× speedup during inference.

Rui Ha, Rui Pu, Chaozhuo Li, Li Sun, Sen Su

Large Reasoning Models (LRMs) can exhibit step-by-step reasoning, reflection, and backtracking, but these behaviors are often unregulated, leading to overthinking. As a result, LRMs continue generating redundant reasoning even after reaching high-confidence conclusions. This increases inference cost and latency, limiting practical deployment. The root cause is the absence of an intrinsic mechanism to monitor the reasoning state and decide when to continue, backtrack, or stop. We propose MERA, a meta-cognitive reasoning framework that decouples reasoning from control to enable independent optimization of control strategies. MERA constructs high-quality reasoning–control supervision data via a takeover-based pipeline, and transforms long-horizon traces into structured reasoning–control alternating sequences for training. The model is trained with supervised fine-tuning to internalize the structured separation, and further optimized with Control-Segment Policy Optimization (CSPO), which combines segment-wise GRPO with control masking to focus learning on control segments. Experiments across reasoning benchmarks show that MERA improves both efficiency and accuracy.

Junhao Jia, Hanwen Zheng, Yueyi Wu, Huangwei Chen, Haishuai Wang, Jiajun Bu, Lei Wu

Natural language data is inherently noisy, yet standard interpretable models often rely on scalar similarities that obscure the true evidentiary basis of a prediction. This limitation is particularly detrimental to prototype-based classification, where traditional full-alignment mechanisms force non-informative background segments to match informative prototypes, yielding unstable or misleading explanations. To mitigate this, we present SCOUT, a novel paradigm that grounds prototype reasoning in the selective correspondence of discriminative fragments. Concretely, we represent each document as a discrete distribution over span embeddings and employ differentiable Unbalanced Optimal Transport (UOT) to align them with class-specific prototypes. Unlike standard methods, this mechanism enables the model to focus strictly on decisive evidence while leaving irrelevant noise unmatched via geometric mass suppression. To ensure verifiability, we anchor prototype supports to readable training spans, establishing a transparent bridge between input segments and stored knowledge. Comprehensive experiments on seven benchmarks demonstrate that SCOUT yields prototypes focused on semantically significant spans, significantly outperforming traditional rationale extraction and post-hoc attribution methods in terms of faithfulness and stability.

Junhoo Lee, Seungyeon Kim, Nojun Kwak

Diffusion Language Models (DLMs) have emerged as a promising alternative to Autoregressive Language Models, yet their inference strategies largely rely on prefix-based prompting inherited from the autoregressive paradigm. In this paper, we propose Template Infilling (TI), a conditioning methodology tailored for DLMs. Unlike conventional prefix prompting, TI distributes structural anchors across the target response, establishing a global template before infilling masked segments. This enables structured conditioning that leverages the bidirectional generation process of DLMs. We evaluate TI on diverse benchmarks, including mathematical reasoning, code generation, and trip planning, achieving consistent improvements of 9.40%p over baseline prompting strategies. Furthermore, TI naturally supports multi-token generation settings, providing practical speed advantages while maintaining generation quality and robustness. Overall, our results highlight a DLM-specific conditioning paradigm for structured generation, suggesting a promising direction for inference methods tailored to diffusion-based language models.

Shaomu Tan, Dawei Zhu, Ke Tran, Michael Denkowski, Sony Trenous, Leonardo F. R. Ribeiro, Bill Byrne, Felix Hieber

Iterative refinement is a simple inference-time strategy for machine translation: given an initial translation, an LLM revises it without additional training. Yet document-scale refinement remains poorly understood: 1) which pipelines work best, 2) what quality dimensions improve, and 3) how refiners behave. In this paper, we present a systematic study of document-level literary translation, covering six LLMs and seven language pairs. Across nine translation-refinement granularity combinations and five refinement strategies, a) we find a robust recipe: document-level MT followed by segment-level refinement yields the strongest and most stable improvements. In our setting, doc-level refinement often makes fewer edits and leads to smaller or less reliable gains. Surprisingly, a simple general refinement prompt consistently outperforms error-specific prompting and evaluate-then-refine schemes. b) Fine-grained MQM analyses and professional-translator evaluation show that gains come primarily from fluency, with limited improvements in adequacy. c) Probing translator-refiner strength interactions suggests refinement behaves less like targeted post-editing and more like projecting outputs toward the refiner’s learned distribution while remaining anchored to the initial translation.

Jiajie Jin, Yuyao Zhang, Yimeng Xu, Hongjin Qian, Yutao Zhu, Zhicheng Dou

Professional financial reports serve as the cornerstone of investment decisions, demanding deep reasoning and multimodal synthesis. While recent deep research systems excel in open-domain search, they struggle with financial reporting, specifically in handling financial data, ensuring analytical depth, and integrating professional visualizations. To address this, we introduce FinSight , the first multi-agent framework for automate end-to-end professional, multimodal financial report. At its core, we propose the Code Agent with Variable Memory architecture, which unifies data, tools, and agents into a programmable variable space, enabling flexible data manipulation and reasoning through executable code. To guarantee report quality, FinSight incorporates a Two-Stage Writing Framework with Generative Retrieval. This mechanism first distills raw data into structured Chain-of-Analysis segments, and then progressively synthesizes them into a coherent, citation-aware, and multimodal narrative. Additionally, an Iterative Vision-Enhanced Mechanism leverages visual feedback to refine code-generated charts to expert standards. Experiments on company and industry-level tasks demonstrate that FinSight significantly outperforms leading deep research systems in factual accuracy, analytical depth, and presentation quality, demonstrating a clear path toward generating professional financial reports. Our code is available at https://anonymous.4open.science/r/FinSight-5841.

Lang Feng, Fuchao Yang, Feng Chen, Xin Cheng, Haiyang Xu, Zhenglin Wan, Ming Yan, Bo An

Recent advances in large language models (LLMs) enable agentic systems trained with reinforcement learning (RL) over multi-turn interaction, but practical deployment is bottlenecked by rapidly growing textual histories that inflate token and memory costs. We introduce AgentOCR, a framework that exploits visual tokens’ superior information density by representing the accumulated observation-action history as a compact rendered image. To make multi-turn rollouts scalable, AgentOCR proposes segment optical caching. By decomposing history into hashable segments and maintaining a visual cache, this mechanism eliminates redundant re-rendering. Beyond fixed rendering, AgentOCR introduces agentic self-compression, where the agent actively emits a compression rate and is trained with compression-aware reward to adaptively balance task success and token efficiency. We conduct extensive experiments on challenging agentic benchmarks, ALFWorld and search-based QA. Remarkably, AgentOCR preserves over 95% of text-based agent performance while substantially reducing token consumption (>50%), yielding consistent token and memory efficiency. Further analysis validates a 20\times rendering speedup from optical caching and effective self-compression balancing. Our code is available at https://github.com/langfengQ/AgentOCR.

Yihang Li, Chenhui Chu

Evaluating meeting effectiveness is crucial for improving organizational productivity. Current approaches rely on post-hoc surveys that yield a single coarse-grained score for an entire meeting. The reliance on manual assessment is inherently limited in scalability, cost, and reproducibility. Moreover, a single score fails to capture the dynamic nature of collaborative discussions. We propose a new paradigm for evaluating meeting effectiveness centered on novel criteria and temporal fine-grained approach. We define effectiveness as the rate of objective achievement over time and assess it for individual topical segments within a meeting. To support this task, we introduce the AMI Meeting Effectiveness (AMI-ME) dataset, a new meta-evaluation dataset containing 2,459 human-annotated segments from 130 AMI Corpus meetings. We also develop an automatic effectiveness evaluation framework that uses a Large Language Model (LLM) as a judge to score each segment’s effectiveness relative to the overall meeting objectives. Through substantial experiments, we establish a comprehensive benchmark for this new task and evaluate the framework’s generalizability across distinct meeting types, ranging from business scenarios to unstructured discussions. Furthermore, we benchmark end-to-end performance starting from raw speech to measure the capabilities of a complete system. Our results validate the framework’s effectiveness and provide strong baselines to facilitate future research in meeting analysis and multi-party dialogue. Our dataset and code will be publicly available.

Yanzhi Tian, Cunxiang Wang, Zeming Liu, Heyan Huang, Wenbo Yu, Dawei Song, Jie Tang, Yuhang Guo

Large Language Models (LLMs) have significantly advanced Machine Translation (MT), applying them to linguistically complex domains-such as Social Network Services, literature etc. In these scenarios, translations often require handling non-literal expressions, leading to the inaccuracy of MT metrics. To systematically investigate the reliability of MT metrics, we first curate a meta-evaluation dataset focused on non-literal translations, namely MENT. MENT encompasses four non-literal translation domains and features source sentences paired with translations from diverse MT systems, with 7,530 human-annotated scores on translation quality. Experimental results reveal the inaccuracies of traditional MT metrics and the limitations of LLM-as-a-Judge, particularly the knowledge cutoff and score inconsistency problem. To mitigate these limitations, we propose RATE, a novel agentic translation evaluation framework, centered by a reflective Core Agent that dynamically invokes specialized sub-agents. Experimental results indicate the efficacy of RATE, achieving an improvement of at least 3.2 points in combined system- and segment-level correlation with human judgments compared with current methods. Further experiments demonstrate the robustness of RATE to general-domain MT evaluation. Code and dataset are available at: https://github.com/BITHLP/RATE.

Shaohua Duan, Pengcheng Huang, Xinze Li, Zhenghao Liu, Xiaoyuan Yi, Yukun Yan, Shuo Wang, Yu Gu, Ge Yu, Maosong Sun

Long-context modeling is critical for a wide range of real-world tasks, including long-context question answering, summarization, and complex reasoning tasks. Recent studies have explored fine-tuning Large Language Models (LLMs) with synthetic data to enhance their long-context capabilities. However, the effectiveness of such approaches is often limited by the low diversity and factual inconsistencies in the generated data. To address these challenges, we propose LongMab, a novel framework that leverages a Multi-Armed Bandit (MAB) rollout strategy to identify the most informative chunks from the given long context for sampling high-quality and diverse responses and constructing preference data pairs for Direct Preference Optimization (DPO) training. Specifically, we treat context chunks as arms of MAB, select chunks based on their expected reward scores to input into LLMs to generate responses, and iteratively update these scores based on reward feedback. Both exploration and exploitation during the rollout process enable the LLM to focus on the most relevant context segments, thereby generating and collecting high-quality and diverse responses. Experimental results on both Llama and Qwen show the effectiveness of LongMab by achieving more than a 4% improvement on long-context reasoning benchmarks. All data and code will be released on https://github.com/NEUIR/LongMab-PO.

Yage Zhang, Yukun Jiang, Michael Backes, Yang Zhang

Anomaly detection (AD) plays a critical role in applications such as automated industrial inspection and medical image analysis. Empowered by the strong pre-trained vision-language model, CLIP, recent years have witnessed the emergence of several CLIP-based few-shot AD methods.Due to the overlap between the embedding distributions of normal and anomalous samples, many existing approaches introduce additional model training for more discriminative text embeddings.However, we demonstrate that such training is not necessary.Specifically, we find that this embedding overlap can be separated by introducing a \underline{\text{Diff}}erence-guided vector for embedding \underline{\text{Edit}}ing (DiffEdit).Based on this finding, we propose DE-CLIP, a simple yet effective framework based on DiffEdit, which directly edits text embeddings based on the textual and visual differences between normal and anomalous samples, resulting in more discriminative embeddings for AD.Extensive experiments on industrial and medical datasets demonstrate the superiority of our proposed DE-CLIP compared with existing baselines.For instance, on MVTec dataset, DE-CLIP achieves 96.6% and 96.7% AUROC on anomaly classification and segmentation, surpassing both training-based and training-free methods.In addition, we observe that introducing DiffEdit into other training-free baselines could also significantly improve their performance, highlighting the potential of DiffEdit to promote better AD.

Siqi Ouyang, Shuoyang Ding, Oleksii Hrinchuk, Vitaly Lavrukhin, Brian Yan, Boris Ginsburg, Lei Li

Simultaneous speech translation (SST) generates translations while receiving partial speech input. Recent advances show that large language models (LLMs) can substantially improve SST quality, but at the cost of high computational overhead. To reduce this cost, prior work reformulates SST as a multi-turn dialogue task, enabling full reuse of the LLM’s key–value (KV) cache and eliminating redundant feature recomputation. However, this approach relies on supervised fine-tuning (SFT) data in dialogue form, for which few human annotations exist, and existing synthesis methods cannot guarantee data quality. In this work, we propose a Hierarchical Policy Optimization (HPO) approach that post-train models trained on imperfect SFT data. We introduce a hierarchical reward that balances translation quality and latency objectives. Experiments on English to Chinese/German/Japanese demonstrate improvements of over +7 COMET score and +1.25 MetricX score at a latency of 1.5 seconds. Comprehensive ablation studies further validate the effectiveness of different quality rewards, hierarchical reward formulations, and segmentation strategies.

Yixiao He, Menghao Zhang, Haifeng Sun, Jing Wang, Kangheng Lin, Jinghan Wang, Chenye Xu, Pengfei Ren, Qi Qi, Jingyu Wang

Video anomaly understanding (VAU) is critical for real-world scenarios. Recent advances in Video Large Language Models (Video-LLMs) enhance the ability of VAU models to describe and interpret anomalies. However, progress in anomaly localization is still limited by two key issues. First, most existing video anomaly datasets only annotate segments that are clearly inconsistent with the context, often omitting subsequent segments that are semantically part of the same abnormal event. Second, the field lacks systematic evaluation protocols. To bridge these gaps, we introduce VALU, a new benchmark that explicitly defines anomalies across five semantic levels and provides comprehensive temporal boundaries and detailed textual descriptions for each. Based on these annotations, we design three evaluation tasks that comprehensively assess models’ capabilities across different dimensions, including temporal grounding, anomaly localization, and anomaly detail discrimination. Evaluation results reveal persistent challenges in current models’ capabilities on VAU. We further analyze and discuss these findings, and hope that both VALU and insights will advance research in VAU and the development of Video-LLMs. Our benchmark will be publicly available.