A deployed LLM coaching agent answers requests whose correct response depends on player state and game version, so advice that is right for one player can be wrong for the next. We describe TactiNavi, the serving system behind the LLM coach of a commercial open-world game, which makes strategy applicability an explicit artifact shared by offline compilation, online execution, failure diagnosis, and release. An offline compiler turns versioned game evidence and expert guidance into strategy packages with explicit prerequisites, the serving plane evaluates those prerequisites as three-valued conditions over structured player state and records decision traces, and a control plane uses the resulting condition-rule dependencies to validate revisions on held-out traffic before release. In a retrospective evaluation of 1,122 logged coaching turns from three deployment periods, condition-bearing packages raise correctness from 57.9% to 82.0% over atomic facts and reduce strategy misapplication under changed player state from 18.1% to 10.4%, while trace-based claim verification reaches 86.4% precision at a median reply latency of 2.07 s. The deployment also showed that most apparent fixes are generation noise, because an unchanged-system rerun recovers 38.4% of failing turns and only one of 15 shipped revision units passes the release checks retrospectively. We distill these results into lessons on treating applicability as a first-class serving object, sizing traces for later diagnosis, and building the null model into the release path.
Coarse-to-fine image-to-point-cloud (I2P) registration risks discarding correct but ambiguous patch pairs, while retaining more candidates repeats fine matching over overlapping patches. SparseGraph-I2P aims to improve registration recall by increasing the inlier ratio at the pose solver, using sensor or monocular depth. A Coarse Hypothesis Graph combines appearance with depth-derived first- and second-order compatibility to select a fixed budget of patch pairs. Sparse Fine Matching decodes each queried pixel once and skips pixel--point comparisons covered by an ancestor patch paired with the same point-cloud region. With sensor depth, depth-aware filtering ranks correspondences by SC2 spectral support for a single EPnP solve. With monocular depth, depth or inverse depth is calibrated from correspondence geometry before filtering. SparseGraph-I2P achieves the highest registration recall and lowest latency among the compared methods on RGB-D Scenes V2 and 7-Scenes with sensor depth. With predicted depth, it retains the highest recall on RGB-D Scenes V2 at an additional calibration cost. Code will be released.
Ruikai Li, Naiyu Gao, Hao Shan, Yizhe Fan, Yanhao Wu, Han Jiang, Yilong Ren, Zhenning Li 等 14 位作者
Real-world driving requires context-aware decision-making. Existing end-to-end driving benchmarks primarily assess trajectory quality through safety, compliance, comfort, and driving progress, leaving context-dependent decision-making insufficiently characterized. To fill this gap, we present PlanContext, a benchmark for evaluating in-context driving decision-making. PlanContext constructs paired evaluations by varying high-level driving intentions while keeping the observed scene fixed, enabling a natural and controllable assessment of context-dependent decision-making. We further define six representative driving contexts and jointly evaluate contextual decision correctness, road compliance, and dynamic safety with class-balanced metrics. Experiments on 9,561 scene-condition instances from NAVSIM and 23 public end-to-end driving models show that high conventional driving-quality scores do not consistently correspond to strong route-level in-context driving decision-making. Further analysis shows that model behavior varies with decision timing, road structure, and traffic interaction depending on the decision required by the current context. Our results highlight in-context driving decision-making as an important complementary dimension for evaluating end-to-end driving systems.
Autoregressive video diffusion models (AR-VDMs) enable real-time, long-horizon video generation but degrade rapidly once the inference horizon outruns the training horizon, due to compounding exposure bias. To stabilize long rollouts, recent methods heuristically adopt frame sink, keeping initial frames as a persistent model context, inspired by attention sinks in LLMs. We revisit this design and find that the mechanism by which initial frames stabilize long-range consistency in AR-VDMs is not an LLM-style attention sink; instead, long-range consistency is anchored through a sparse subset of attention heads, while the majority of heads predominantly capture local temporal dynamics. This observation reveals an over-conditioning pathology in existing frame-sink strategies: by universally exposing every head to the pinned initial-frame context, the model becomes prone to collapsing motion toward the low-dynamic modes of the teacher during model distillation. We propose Disen-Forcing, a distillation-time plug-in for AR-VDMs that disentangles and reinforces this head specialization through disentangled context routing and motion-disentangling regularization. On long-horizon video generation benchmarks, Disen-Forcing simultaneously improves long-range consistency and motion dynamics of base AR-VDMs, a trade-off that sink-based methods fail to balance.
Haoyang Yuan, Boyang Li, Yingqian Wang, Yimian Dai, Nuo Chen, Xinfei Huang, Shuqi Yi, Zaiping Lin 等 10 位作者
Omni-domain infrared small target (IRST) detection is crucial for infrared surveillance, yet remains challenging due to heterogeneous imaging domains and inconsistent target characteristics. Previous deep learning-based methods have been developed for visual-only paradigms and achieved promising performance on domain-specific tasks. However, existing methods follow the task-specific supervised learning paradigm. This paradigm simplifies the full-scene infrared observations to sparse target supervision, discarding the semantics that remain invariant across heterogeneous domains. Consequently, detection performance suffers substantially under domain shifts. To handle this issue, we introduce \textbf{``understand before detect''}, a paradigm that formulates omni-domain IRST detection as an understanding-driven process, where holistic infrared target understanding precedes precise detection. Building on this paradigm, we propose \textbf{JinSight}, which first develops holistic IRST understanding through language supervision and then transfers the learned cross-domain representations to precise small-target detection. By grounding infrared representations in language semantics, JinSight enables a single model to generalize across heterogeneous infrared domains. We then introduce Latent Semantic Interaction (LSI), which exchanges language-aligned global semantics with fine-grained spatial features in a compact low-rank space. To address the lack of multimodal omni-domain IRST benchmarks, we build \textbf{OmniIRST-VL}, the first large-scale, highly diverse vision--language dataset for omni-domain IRST detection. It comprises over 39k annotations across six complementary instruction tasks covering both scene-level understanding and target-centric reasoning.
Guancheng Pan, Qixin Tan, Yang Yuan, Andrew Chi-Chih Yao
AI 速览：提出代数世界模型（AWM），将未知环境中的探索转化为可执行假设、范畴论引导的结构化模型，并在 Lean 4 中形式化证明可达性与不可能性。
AI has made remarkable progress in mathematics, solving challenging problems with machine-checkable proofs in Lean, yet typically relying on predefined mathematical definitions and formal specifications. This raises a natural question: can it reason rigorously about an unknown world? We introduce the Algebraic World Model (AWM), a framework that turns interaction with unfamiliar environments into explicit mathematical models supporting formal reasoning. AWM explores the environment and develops executable Python hypotheses, continually testing and revising them against observations. It then structures these hypotheses into typed objects and shared rules, simplifies their structure through category-theory-guided transformations, and formalizes the resulting model in Lean 4 to prove reachability and impossibility. AWM can formally prove that a goal is unreachable, even beyond explored trajectories. On two impossible game levels, it constructs closed, kernel-checked Lean proofs excluding every legal winning sequence in all six runs, while Claude Opus 5.5 and GPT-6 Astra, equipped with coding tools, produce none in twelve baseline runs. Four of AWM's six initial bounds are correct, compared with one for Opus 5.5 and none for GPT-6 Astra; given only feedback that a cheaper win exists, AWM revises its rules and constructs new proofs, reaching six of six correct bounds. Moreover, AWM produces interpretable mathematical models whose rules can be inspected and tested step by step. Across all 25 public ARC-AGI-3 games, these models enable Lean to replay the winning trajectories of all 183 levels and generalize beyond observed play, predicting unseen transitions and detecting changes in game mechanics. With Claude Opus 5, AWM clears all levels with a score of 99.9, while its certified winning trajectories use only 43% of the human baseline's actions.
ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions
Xianzhong Ding, Yangyang Yu, Changwei Liu, Bill Zhao, Le Chen, Tao Chen
AI 速览：作者提出 ContextEcho 基准与可复用测试框架，用于刻画长程智能体编程会话中编程助手人格的漂移现象。
Frontier language models are trained to exhibit stable programming assistant personas, yet those personas may drift over long agentic coding sessions in production. After a long real-world coding session, a model that hedges preferences in a fresh chat ("I don't have preferences") may instead express them directly ("Python -- the feedback loop is instant..."). Existing persona-stability studies focus on scripted chat of tens of turns, leaving long production coding sessions involving thousands of tool-using agent steps, repeated compaction, and hours of interaction largely uncharacterized. We introduce ContextEcho, a benchmark and reusable harness with a 25-probe identity suite, a snapshot-then-probe protocol that evaluates conversation state without perturbing the main session, judge-based and judge-free metrics, and a corpus of 57 long agentic-coding sessions collected with consent and redacted for release across Claude Code and Codex CLI. We use one representative session for a 23-model cross-provider study and all 57 sessions for corpus-scale trajectory analysis across frontier and open-weight models. Results show that persona drift occurs across providers and emerges early in real agentic interaction, while in-session compaction neither reliably resets nor amplifies it. A single-shot anchor restores the original assistant persona, though placebo controls suggest a generic register-reset effect. On same-task continuations, drifted prefixes improve tool-call fidelity -- which we attribute to retained task context rather than drift itself. In tool-free chat, however, the same prefixes break output-format contracts and inflate output length. These findings motivate ContextEcho as an open-source framework for auditing persona drift in long coding sessions without retraining.
Multi-hop Knowledge Base Question Answering (KBQA) aims to find answer entities in the knowledge base that are multiple hops away from the entities in the question. Information retrieval-based (IR-based) methods extract a pivotal subgraph from the entire KB to locate candidate answers and then evaluate their plausibility through semantic matching with the question. However, we observed that the extracted subgraphs often include nodes that are weakly related or irrelevant to the question. Without a proper node filtering mechanism, the number of irrelevant nodes grows as the number of hops increases, leading to excessive consumption of computational resources. To address these challenges, this study introduces an efficient LLM-based subgraph retrieval method for multi-hop knowledge base question answering, M-ER. The framework leverages Monte Carlo Tree Search (MCTS) to transform subgraph exploration into a tree-structured search process. During the MCTS selection phase, nodes that are highly relevant to the question are prioritized for inclusion in the subgraph, eliminating the need to traverse all nodes in the KB. The framework further incorporates a large language model (LLM) to refine the search direction, ensuring that exploration remains focused on nodes relevant to the question. In addition, selected nodes are quantitatively scored, and these scores are fed back into the node selection process to effectively filter out irrelevant candidates, thereby improving the quality of the subgraph. This mechanism not only narrows the search space but also enhances the overall efficiency of multi-hop KBQA. Experiments on the WebQSP benchmark demonstrate that M-ER achieves 78.88\% on the Hits@1 metric, while also improving computational efficiency. These results not only validate the effectiveness of M-ER, but also offer a viable technical path to balance performance and computational efficiency.
Chenyang Yu, Zhaomin Xiao, Yan Huang, Eduardo Blanco
Gardening in the backyard implies living there, while checking into an Airbnb more likely means staying for a few days. Such duration judgments are important for understanding human mobility, travel behavior, and social context. However, duration is rarely stated explicitly in social media posts and must instead be inferred from noisy textual, visual, and contextual cues. In this paper, we investigate whether vision-language models possess such duration commonsense. We create a new corpus of geotagged tweets co-annotated by humans and LLMs to serve as a testbed for this task. We find that state-of-the-art models struggle with duration inference, even when guided by multi-turn Chain-of-Thought prompting. Error analyses further reveal that models often fail to integrate critical contextual information, or they overly rely on prominent visual details while overlooking subtle hashtags and semantic nuances. These findings underscore the significant challenges remaining in context-aware multimodal understanding.
We introduce Nemotron 3.5 Super and Nemotron 3.5 Lightning, open and efficient Mixture-of-Experts Hybrid Mamba-Transformer models designed for reasoning, long-context understanding, and agentic workloads. Super is a 120B-parameter, 12B-active multimodal model, while Lightning is a smaller 30B-parameter, 3B-active text model; both designed for efficient deployment. Both models combine large-scale distillation, long-context training, and native speculative decoding, while Super further incorporates Indic continued pretraining and agentic software-engineering mid-training. We develop a comprehensive post-training pipeline spanning supervised fine-tuning and reinforcement learning (RLVR and RLHF), complemented in Super by specialized teacher training and multi- teacher on-policy distillation (MOPD). This transforms Super from a text-only base model into a multimodal agent capable of reasoning over text, images, video, and interactive tool environments. We further explore test-time scaling through parallel agent execution and self-verification. Across training and deployment, we jointly optimize capability and efficiency. Indic continued pretraining improves Indian-language benchmarks by up to 9.82 percentage points, while agentic mid-training strengthens software-engineering capabilities and robustness across agent harnesses. The post-trained Super model achieves 51.78% on SWE-bench Pro, surpassing Nemotron 3 Ultra despite using approximately 4.6x fewer active parameters. Post-training MTP boosting improves speculative- decoding acceptance length by up to 21.8% under probabilistic draft sampling, while NVFP4 quantization achieves 99.62% median accuracy recovery relative to BF16. Inference optimizations enable up to 4,856 output tokens/s/GPU on GB200 while sustaining at least 50 tokens/s/user.
Yipeng Hao, Yuxuan Cai, Wenkai Ji, Jie Zhou, Liyang Yu, Qin Chen, Zhiguo Yan, Shulei Zheng 等 9 位作者
Memory helps LLM-based multi-agent systems reuse experience, but stored advice can be flawed or applied in the wrong situation. Team-level success or failure alone cannot reliably show whether a memory needs correction or should not have been used. We propose DiscernMem, a training-free framework that improves both memory quality and memory use. To improve quality, Defect-Aware Memory Refinement (DMR) compares team executions with a single memory and with no memory, diagnoses rule defects, and tests repairs on current and earlier tasks to guard against losing previously useful behavior. To guide appropriate use, Boundary-Learning Meta-Memory (BLM) records each memory's past uses and effects, learning when its owner role should apply or avoid it based on task context and incoming messages. Together, these components help teams correct flawed experience while retaining useful rules that were applied incorrectly. Experiments with Qwen3-4B-Instruct-2507 on four benchmarks show that DiscernMem achieves the highest average score on AutoGen, CAMEL, and MacNet, exceeding the strongest reported baselines by 0.51-4.32 points. It leads on code generation and symbolic planning across all three systems, including a PDDL score of 37.71 versus 28.48 on CAMEL.
Generalist robot policies and language-model agents provide complementary capabilities, but completing a task also requires adapting their use to the consequences of physical interaction. We study execution as a locus of adaptation: feedback guides the current action and supplies evidence for revising subsequent decisions under fixed model weights. We introduce RoboDoctor, a training-free framework that connects diagnosis, grounded action, and persistent experience across three revision targets. The harness allocates control between policy and agent; intervention determines when to assist while preserving useful policy behavior; and operation refines physical methods through observed effects. Linked experience records retain the reason for a decision, the attempted method, its outcome, and conditions for reuse. On a curated archive of 54 RoboDojo-Sim task entries, RoboDoctor achieves a five-dimension mean score of 57.10, with a native success rate of 49.44%. Recorded interactions illustrate strategy changes, selective correction, and grasp revision. These findings characterize execution-time adaptation in the evaluated system; controlled trials remain necessary to isolate the contribution of cross-episode experience.
While unmanned aerial vehicles (UAVs) are increasingly deployed for pipeline inspection, prevailing research predominantly assumes static pipes. In real-world industrial environments, pipes often exhibit dynamic movements, such as vibrations. This imposes more stringent requirements on UAVs' adaptive control and computational efficiency than in static scenarios. Therefore, this paper proposes a servo control framework named VisGeo-PipeTrack, which utilizes geometric constraints to enable UAV tracking of dynamic pipes. To achieve efficient UAV positioning, YOLOv5 is employed to detect labels on the pipe. The positions are derived from visual geometric constraints based on these detected labels to enable precise localization. To address tracking control challenges, a model predictive control based real-time corrected adaptive PID (MPC-RCAPID) controller is proposed. This controller employs the Recursive Least Squares method for real-time model estimation and integrates rolling horizon optimization to dynamically adjust PID parameters. Experiments conducted on both the AirSim simulator and the Tello UAV platform validate the effectiveness of the proposed framework. Simulation comparisons with standard PID, fractional-order PID, adaptive PID, and fuzzy PID demonstrate that MPC-RCAPID achieves the shortest settling time (3.26 s) and the lowest average tracking error (3.52%). In physical experiments on the Tello platform, compared with standard PID control, MPC-RCAPID reduces the average tracking error by 55.52%, 26.72%, and 26.52% across the three test dimensions. These results demonstrate that VisGeo-PipeTrack facilitates efficient autonomous tracking in dynamic pipe environments.
Shuyang Yu, Haibo Qin, Shaohan Hu, Wei Shan, Xiaoqing Guan, You Wang, Guang Li
Monocular occupancy prediction in indoor scenes aims to recover a scene's 3D geometry and semantics from a single RGB image. Existing methods are restricted to a closed set of predefined categories, limiting their ability to generalize to unseen categories. This limitation is particularly critical in indoor scenes, where objects exhibit highly diverse and fine-grained semantic categories. We study open-vocabulary monocular occupancy prediction for indoor scenes and present OVM-Occ, a Gaussian-based occupancy framework with an instance-aware vision-language teacher. Specifically, we adopt Language-Embedded Gaussians as the intermediate scene representation, propose a geometry-aware deformable attention module to refine Gaussian semantic features, and construct instance-level vision-language embeddings to provide reliable semantic supervision. We further introduce geometry-aware decoupled semantic distillation to mitigate the interference between geometry learning and semantic learning. To evaluate open-vocabulary occupancy prediction performance, we introduce OpenOcc-ScanNet and OpenOcc-Replica, two extended benchmarks with more diverse and fine-grained semantic categories. Experiments show that our method achieves state-of-the-art performance in open-vocabulary monocular occupancy prediction for indoor scenes, outperforming the previous state-of-the-art approach by 12.2 mIoU and 0.6 IoU on the Occ-ScanNet dataset, with consistent gains under new open-vocabulary evaluation protocols.
Estimating the correctness of large language model (LLM) predictions without access to ground truth is critical for reliable deployment in high-stakes settings. Existing approaches, such as entropy and perplexity, operate solely on final output statistics, discarding the rich information encoded in intermediate reasoning steps. To address this limitation, we propose STEP (Soft Trajectory Estimation of Prediction correctness), a training-free framework that scores each reasoning step via lightweight forward passes to obtain a soft probability distribution qi over candidate answers, from which we derive a set of soft trajectory features capturing per-step confidence and its progression toward the final answer. Extensive experiments on various benchmarks demonstrate consistent and substantial improvements, particularly in AUROC, over output-based baselines and hard-indicator trajectory methods. Notably, Temporally-Weighted Consistency, which assigns higher importance to later reasoning steps where answer distributions have stabilized, contributes the largest gains, highlighting the importance of temporal structure in reasoning trajectories. These results further validate the advantage of preserving continuous probability signals over binary discretization.
Justice Ou University of Illinois at Urbana-Champaign, Tinglin Huang, Yilun Zhao Yale University, Ziyang Yu, Peiqing Lu Microsoft, Yifei Shen, Rex Ying Yale University
To improve the reliability of Large Language Models (LLMs) in clinical applications, retrieval-augmented generation (RAG) is extensively applied to provide factual medical knowledge. However, beyond general medical knowledge from open-ended datasets, clinical case-based knowledge is also critical for effective medical reasoning, as it provides context grounded in real-world patient experiences. Motivated by this, we propose Experience Retrieval-Augmentation ExpRAG framework based on Electronic Health Record(EHR), aiming to offer the relevant context from other patients’ discharge reports. ExpRAG performs retrieval through a coarse-to-fine process, utilizing an EHR-based report ranker to efficiently identify similar patients, followed by an experience retriever to extract task-relevant content for enhanced medical reasoning. To evaluate ExpRAG, we introduce DischargeQA, a clinical QA dataset with 1,280 discharge-related questions across diagnosis, medication, and instruction tasks. Each problem is generated using EHR data to ensure realistic and challenging scenarios. Experimental results demonstrate that ExpRAG consistently outperforms a text-based ranker, achieving an average relative improvement of 5.2%, highlighting the importance of case-based knowledge for medical reasoning.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Guanglu Sun Harbin University of Science and Technology, Xinyu Liu, Lili Liang Harbin University of Science and Technology, Yang Yu Harbin University of Science and Technology, Fei Lang Harbin University of Science and Technology, Suxia Zhu, Ming Liu Harbin Institute of Technology
In real-world scenarios, multimodal information continuously evolves, with new entity and relation types emerging, necessitating timely updates to multimodal knowledge graphs for supporting downstream tasks. However, existing methods struggle to balance real-time adaptability and computational efficiency in continual learning scenarios. To this end, this paper proposes the Continual Multimodal Entity and Relation Joint Extraction (CMERJE) task and a Multimodal Prompt-based Boundary-enhanced Continual (MPBoCo) framework. Specifically, MPBoCo incrementally stores task-specific knowledge via learnable multimodal prompts, dynamically matches relevant prompts for each instance, and fuses them into a frozen backbone model for task-specific reasoning. Subsequently, the boundary-enhanced dual-branch module leverages the auxiliary branch to preserve local syntactic continuity and provide boundary guidance. Experimental results demonstrate that MPBoCo achieves superior performance in real-world scenarios, significantly outperforming baseline methods by 5.5% and 7.2% in 10-task and 5-task settings, respectively.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Haibo Tong, Zeyang Yue, Feifei Zhao Institute of automation, Chinese academy of science, Chinese Academy of Sciences, Erliang Lin, Lu Jia, Ruolin Chen Institute of automation, Chinese academy of science, Chinese Academy of Sciences, Yinqian Sun Institute of automation, Chinese academy of science, Chinese Academy of Sciences, Qian Zhang Institute of automation, Chinese academy of science, Chinese Academy of Sciences 等 9 位作者
Whether Large Language Models (LLMs) truly possess human-like Theory of Mind (ToM) capabilities has garnered increasing attention. However, existing benchmarks remain largely restricted to narrow paradigms like false belief tasks, failing to capture the full spectrum of human cognitive mechanisms. We introduce CogToM, a comprehensive, theoretically grounded benchmark comprising over 8000 bilingual instances across 46 paradigms, validated by 49 human annotators. A systematic evaluation of 22 representative models, including frontier models like GPT-5.1 and Qwen3-Max, reveals significant performance heterogeneities and highlights persistent bottlenecks in specific dimensions. Further analysis based on human cognitive patterns suggests potential divergences between LLM and human cognitive structures. CogToM offers a robust instrument and perspective for investigating the evolving cognitive boundaries of LLMs. We release our code and data at https://github.com/Beijing-AISI/CogToM.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
YongKang Liu Northeast University, Jiayang Yu Peking University, Mingyang Wang LMU Munich, Yiqun Zhang Shanghai Artificial Intelligence Laboratory and Northeastern University, Ercong Nie Shanghai Jiaotong University, Shi Feng Northeastern University, China, Daling Wang, Kaisong Song 等 9 位作者
Argumentation generation has attracted substantial research interest due to its central role in human reasoning and decision-making. However, most existing argumentative corpora focus on non-interactive, single-turn settings, either generating arguments from a given topic or refuting an existing argument. In practice, however, argumentation is often realized as multi-turn dialogue, where speakers defend their stances and employ diverse argumentative strategies to strengthen persuasiveness. To support deeper modeling of argumentation dialogue, we present the first large-scale Strategic Argumentative Dialogue dataset, SAD, consisting of 392,822 examples. Grounded in argumentation theories, we annotate each utterance with five strategy types, allowing multiple strategies per utterance. Unlike prior datasets, SAD requires models to generate contextually appropriate arguments conditioned on the dialogue history, a specified stance on the topic, and targeted argumentation strategies. We further benchmark a range of pretrained generative models on SAD and present in-depth analysis of strategy usage patterns in argumentation.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Guibin Zhang National University of Singapore, Haiyang Yu, Kaiming Yang, Bingli Wu, Fei Huang Alibaba Group US, Yongbin Li Alibaba Group, Shuicheng Yan National University of Singapore
Complex agentic AI systems, powered by a coordinated ensemble of Large Language Models (LLMs), tool and memory modules, have demonstrated remarkable capabilities on intricate, multi-turn tasks. However, this success is shadowed by prohibitive economic costs and severe latency, exposing a critical, yet underexplored, trade-off. We formalize this challenge as the Agent System Trilemma: the inherent tension among achieving state-of-the-art performance, minimizing monetary cost, and ensuring rapid task completion. To dismantle this trilemma, we introduce EvoRoute, a self-evolving model routing paradigm that transcends static, pre-defined model assignments. Leveraging an ever-expanding knowledge base of prior experience, EvoRoute dynamically selects Pareto-optimal LLM backbones at each step, balancing accuracy, efficiency, and resource use, while continually refining its own selection policy through environment feedback. Experiments on challenging agentic benchmarks such as GAIA and BrowseComp+ demonstrate that EvoRoute, when integrated into off-the-shelf agentic systems, not only sustains or enhances system performance but also reduces execution cost by up to 80% and latency by over 70%.
尚无 AI 解读，展示原文摘要
cs.CL · cs.AI
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Dong Shu Northwestern University, Northwestern University, Haoyang Yuan NewsBreak, Yuchen Wang, Yanguang Liu, Huopu Zhang, Mengnan Du The Chinese University of Hong Kong, Shenzhen
Large vision-language models (LVLMs) have made significant progress in chart understanding. However, financial charts, characterized by complex temporal structures and domain-specific terminology, remain notably underexplored. We introduce FinChart-Bench, the first benchmark specifically focused on real-world financial charts. FinChart-Bench comprises 1,200 financial chart images collected from 2015 to 2024, each annotated with True/False (TF), Multiple Choice (MC), and Question Answering (QA) questions, totaling 7,016 questions. We conducted a comprehensive evaluation of 26 state-of-the-art LVLMs on FinChart-Bench. Our evaluation reveals critical insights: (1) the performance gap between open-source and closed-source models is narrowing, (2) performance degradation occurs in upgraded models within families, (3) many models struggle with instruction following, (4) both advanced models show significant limitations in spatial reasoning abilities, and (5) current LVLMs are not reliable enough to serve as automated evaluators. These findings highlight important limitations in current LVLM capabilities for financial chart understanding.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Xueqing Peng Yale University, Lingfei Qian Yale University, Yan Wang Yale University, Ruoyu Xiang, Yueru He Columbia University, Yang Ren Yale University, Mingyang Jiang, Vincent Jim Zhang 等 47 位作者
Real-world financial analysis involves information across multiple languages and modalities, from reports and news to scanned filings and meeting recordings. Yet most existing evaluations of LLMs in finance remain text-only, monolingual, and largely saturated by current models. To bridge these gaps, we present MultiFinBen, the first expert-annotated multilingual (five languages) and multimodal (text, vision, audio) benchmark for evaluating LLMs in realistic financial contexts. MultiFinBen introduces two new task families: multilingual financial reasoning, which tests cross-lingual evidence integration from filings and news, and financial OCR, which extracts structured text from scanned documents containing tables and charts. Rather than aggregating all available datasets, we apply a structured, difficulty-aware selection based on advanced model performance, ensuring balanced challenge and removing redundant tasks. Evaluating 21 leading LLMs shows that even frontier multimodal models like GPT-4o achieve only 46.01% overall, stronger on vision and audio but dropping sharply in multilingual settings. These findings expose persistent limitations in multilingual, multimodal, and expert-level financial reasoning. All datasets, evaluation scripts, and leaderboards are publicly released.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Yang Yue, Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang
Reinforcement Learning with Verifiable Rewards (RLVR) has recently demonstrated notable success in enhancing the reasoning performance of large language models (LLMs), particularly on mathematics and programming tasks. Similar to how traditional RL helps agents explore and learn new strategies, RLVR is believed to enable LLMs to continuously self-improve, thus acquiring novel reasoning abilities beyond those of the corresponding base models. In this study we critically examine the current state of RLVR by systematically probing the reasoning capability boundaries of RLVR-trained LLMs across various model families, RL algorithms, and math, coding, and visual reasoning benchmarks, using pass@k at large k values as the evaluation metric. Surprisingly, we find that the current training setup does not elicit fundamentally new reasoning patterns. While RLVR-trained models outperform their base models at small k (e.g., k = 1), the base models achieve a higher pass@k score when k is large. Coverage and perplexity analyses show that the observed reasoning abilities originate from and are bounded by the base model. Treating the base model as an upper bound, our quantitative analysis shows that six popular RLVR algorithms perform similarly and remain far from optimal in leveraging the potential of the base model. By contrast, we find that distillation can introduce new reasoning patterns from the teacher and genuinely expand the model's reasoning capabilities. Overall, our findings suggest that current RLVR methods have not yet realized the potential of RL to elicit truly novel reasoning abilities in LLMs. This highlights the need for improved RL paradigms, such as continual scaling and multi-turn agent-environment interaction, to unlock this potential.