Tianxing Zhu, Huacheng Lu, Ming Liu, Malte Hillebrand
We propose goal-driven self-recursive optical research in which agents use physical evidence to revise the hardware, software, models, and protocols used to pursue a scientific goal. The framework can be instantiated with different stimulus systems and measurement devices. We formalize its updates and derive a relation between physical attainability, command access, and model approximation error. We demonstrate the concept in an optical see-through color-matching case study using an iPad, RayNeo X3 glasses, and a TOP 300 probe coupled to a CAS140D spectroradiometer, both from Instrument Systems. A revised glasses application adds USB-controlled RGB holds and frame acknowledgements, keeping model-generated colors displayed during spectral acquisition. The interface passes 12 live control checks. Spectral feedback reduces one target color difference from 19.756 to 1.211 ΔE_76^*; all five targets improve in a separate comparison. These measurements demonstrate the concept in one configuration. The resulting software and protocols remain available for subsequent investigations.
Han Zhang, Sicheng Lu, Aleks Santari, Jeffrey K. Jopling, Masaru Ishii, Mathias Unberath
Operating rooms (ORs) increasingly contain ambient sensing infrastructure to enable AI-based optimization of workflows and utilization. We pursue an alternative path: using ambient AI to guide robot action and optimize utilization. Deploying humanoids in the OR without ambient AI is severely limited by cumbersome de novo orientation and planning due to partial onboard observations. We posit that this can be achieved by coupling an ambient AI-based agentic orchestrator with the humanoid robotic actor, providing shared perception and execution feedback for humanoid autonomy in the OR. ActOR fuses a wall-mounted ambient camera and the robot's head-mounted camera into a shared object-level digital twin (DT). A large language model agent plans over the DT using pre-programmed locomotion and manipulation skills without task-specific training. During execution, DT updates and execution outcomes inform retries, alternative approaches, or replanning. We evaluated ActOR in simulated and physical environments. Across three tasks in two simulated ORs reconstructed from real rooms, ambient perception reduced exploration from 1.05 to 0.12 Explore calls per trial when targets were initially outside the robot's view. Closed-loop feedback increased success from 46.7% to 83.3% under nominal execution and from 8.3% to 51.7% under injected navigation, object-relocation, and grasp perturbations. In a physical mock OR with a humanoid, ActOR completed 14 of 20 single-object disposal trials and demonstrated replanning after scene changes. Ambient AI-based observations provide scene state that improves robot action planning, increasing success and reducing exploration compared with robot-only observations. Feeding skill outcomes and DT updates back to the planner improves multi-step task success. Our results show the feasibility of humanoid task autonomy for multi-step disposal tasks in a mock OR.
Multi-teacher on-policy distillation consolidates expert capabilities by supervising student-generated trajectories, whereas task arithmetic offers a training-free alternative through linear parameter combinations. We view the linear combination of student and expert parameters as a single differentiable model and propose P$^2$-MOPD, which jointly optimizes the mixing coefficients and the student weights under the same on-policy distillation objective. Unlike task arithmetic, which stays within the span of the frozen student and expert parameters for any choice of merging coefficients, P$^2$-MOPD keeps the student trainable, so the experts act as parameter priors without confining it to this span. Across four domains at 1.7B and 4B scales, P$^2$-MOPD achieves faster and more balanced recovery than MOPD from the same initialization, surpassing its final average recovery in half the training steps. At the end of training, our 1.7B model reaches 94.7% average recovery, versus 80.9% for MOPD.
Coding-agent services are priced by the model they claim, making the verification of such model claims crucial for their users. In this paper, we reveal that existing model auditing methods rely on a latent assumption that the deployment is static, i.e., the claimed model is always served in one reference configuration. This assumption does not hold for coding-agent services, whose providers may legitimately change the sampling preset, the reasoning mode or the system prompt without changing the model. As a result, existing methods either falsely accuse honest providers or miss cheap derivatives of the claimed model. To address this problem, we propose a certified model-claim verification method (dubbed CertClaim). Instead of testing the service against one reference deployment, we make the authorized deployments part of the claim and test whether some authorized deployment explains all the evidence of the service, by calibrating multiple evidence channels in every authorized deployment and fusing them within each deployment. We provide theoretical analyses of its false-accusation rate and detection condition. Extensive experiments on eight claims, two agent harnesses and three task suites verify the effectiveness of CertClaim, which is valid in every setting while separating substitutes far better than existing methods, as well as its resistance to potential adaptive attacks.
As network traffic becomes increasingly dynamic and bursty, conventional centralized traffic engineering (TE) suffers from long control-loop delays caused by network-state aggregation, centralized computation, and policy distribution. In this paper, we propose MATE, a multi-agent TE method for proactive distributed scheduling in hybrid networks with partial deployment of Segment Routing over IPv6 (SRv6). MATE uses historical traffic stability to characterize the burst risk of each traffic demand and path sensitivity to quantify the ability of candidate paths to accommodate additional traffic, thereby enabling differentiated proactive scheduling tailored to heterogeneous traffic characteristics. MATE adopts an offline-centralized training and online-distributed execution architecture. During training, a shared global critic evaluates joint actions under a reward combining maximum link utilization (MLU) with variability-weighted path sensitivity, while counterfactual difference rewards provide agent-specific credit for actor learning. During online execution, edge agents generate traffic-splitting policies from local observations and apply them through SRv6, avoiding centralized optimization in each scheduling interval. Experiments on the China Environment for Network Innovations (CENI) testbed and simulations on the Sanet topology show that MATE outperforms a multi-agent reinforcement learning (MARL) ablation baseline without agent-specific credit assignment, COPE, and TEAL in most evaluated scenarios and approaches centralized FIGRET on measured traffic traces. On CENI-Abilene, MATE achieves MLU within 10% of the linear-programming optimum in up to 90.1% of test instances. MATE also achieves lower median normalized MLU than FIGRET in most tested failure scenarios, with mean neural-inference times of 2.05--3.14~ms.
Yubin Kim, Viresh Pati, Jevon Twitty, Vinh Pham, Shihao Yang, Jiecheng Lu
Transformer architectures have established strong baselines in time series forecasting, yet they typically rely on positional encodings that assume uniform, index-based temporal progression. However, real-world systems, from shifting financial cycles to elastic biological rhythms, frequently exhibit "time-warped" dynamics where the effective flow of time decouples from the sampling index. In this work, we first formalize this misalignment and prove that rotary position embedding (RoPE) is mathematically incapable of representing non-affine temporal warping. To address this, we propose Symplectic Positional Embeddings (SyPE), a learnable encoding framework derived from Hamiltonian mechanics. SyPE strictly generalizes RoPE by extending the rotation group $\mathrm{SO}(2)$ to the symplectic group $\mathrm{Sp}(2,\mathbb{R})$, modulated by a novel input-dependent adaptive warp module. By allowing the attention mechanism to adaptively dilate or contract temporal coordinates end-to-end, our approach captures locally varying periodicities without requiring pre-defined warping functions. We implement this mechanism in StretchTime, a multivariate forecasting architecture that achieves state-of-the-art performance on standard benchmarks, demonstrating superior robustness on datasets exhibiting non-stationary temporal dynamics.
Attention kernels on Tensor Processing Units (TPUs) often underutilize their matrix units (MXUs) because of repeated matrix operand loads, vector-intensive computation, and dependencies that leave MXUs idle. We present FlyWheel, a set of fast TPU kernels for both softmax and linear attention. FlyWheel reformulates softmax and triangular inversion to reduce vector work, while coordinating data layouts to improve matrix operand reuse. It then overlaps independent computation to hide execution latency. We apply these techniques to both softmax attention and Gated DeltaNet on TPU v6e and v7x. FlyWheel speeds up softmax attention by up to $4.14\times$ over the state-of-the-art RPA kernel and $1.80\times$ over Splash Attention, reaching 81% of peak BF16 throughput on v6e. It also speeds up Gated DeltaNet prefill by up to $3.61\times$. Integrated into vLLM serving Qwen3.8-27B, FlyWheel improves prefill throughput by up to $1.56\times$ while maintaining comparable accuracy.
AI 速览：提出自策略状态训练方法，让链式嵌入检索器在自身产生的错误状态上训练，从而更完整地检索多跳证据。
Complex questions in retrieval-augmented generation require evidence spread over several documents, and an answer is reliable only when the retriever finds all of it. Chain embedding retrievers collect such evidence step by step, re-encoding the question together with the passages retrieved so far, and we find that they are already strong: trained only on gold histories, a small chain embedding retriever that calls no language model outperforms graph-based systems and is competitive with a search agent. This training, however, leaves a gap. At test time the retriever conditions on its own retrievals, which soon contain mistakes: by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains. We propose on-policy state training, which trains the encoder on the states it reaches itself. At every epoch the current encoder retrieves chains for the training questions; each state it visits, including those built on its own mistakes, is labeled with the earliest gold passage still missing while the other missing gold passages are masked, so that no correct continuation is penalized; and the encoder is updated contrastively on these states mixed with teacher-forced ones, so that its training distribution follows it as it learns. The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare, and it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct. On the same states, it recovers from its own mistakes far more often, and controlled comparisons attribute the gain to training on the states the encoder currently reaches.
Yueqi Zhu, Yan Pan, Chufan Rui, Jiasheng Luo, Bo Zhou
In safety-critical scenarios, the protection level of the autonomous navigation system is crucial for enabling mobile robots to perform safe tasks. However, existing studies on probabilistic navigation systems for robots usually perform offline accuracy evaluations using limited datasets and assume that the results can be applied to unknown real-world environments. As a result, current autonomous mobile robots often lack protection levels for online safety assessment. To fill this gap, we propose a safety-critical LiDAR-inertial odometry (LIO) that provides deterministic protection levels based on a novel on-manifold set-membership filter. By adopting the unknown but bounded assumption, we derive a neat closed-form relationship between point cloud noise and the uncertainty of the estimation from the iterative closest point algorithm. Using this relationship, we design an on-manifold ellipsoidal set-membership filter (OM-ESMF) and implement it within the LIO system. Leveraging the properties of the set-membership filter, our system offers the feasible sets of the estimated locations as the deterministic protection levels, serving as safety references for the robots' downstream autonomous operations. The experimental results show that our system can provide effective deterministic online safety references for diverse robots in various environments.
Zhefan Rao, Meng Chu, Xuan Billy Zhang, Jingfan Chen, Guohao Cai, Runyi Li, Shiyi Du, Xu Huang 等 92 位作者
World Models (WMs) can support Multimodal Agents (MAs) through simulated experience, while Agent trajectories can inform World Model refinement. Evaluating this interaction requires separate evidence about action selection and state evolution. We introduce Dual Exam, which tests Agents under fixed environment response mechanisms and World Models under fixed action interventions across physical, digital, social, and scientific domains. The Agent track compares 10 configurations on 1,088 items in 15 reporting categories. The World Model track contains 400 cases across eight scenarios; 15 configurations have full coverage and one covers six scenarios. GPT 6 Astra leads the case-weighted Agent Index at 31.3, and GPT-Image-2 leads the full-coverage WM Overall at 21.7. These scores use different tasks and metrics and are not directly comparable. A 150-case pilot of one coupled pairing records zero task successes and motivates component-level diagnosis. Results reveal uneven task performance and failures to preserve task-relevant state. Their value for predicting coupled success or sustained reciprocal learning remains to be established.
In finance, the answer that matters is usually the current one: prices, rates, and disclosures hold only as of a given day, so retrieving up-to-date facts is a core capability of a financial agent. Yet most financial benchmarks overlook timeliness. They fix each answer at annotation time, so the gold goes stale as markets move and, once leaked into training data, rewards recall rather than search and analysis. We introduce LiveSearch, a live benchmark that stores each question as executable verifier code: a typed program of calls to a professional financial data service, followed by a declarative extractor. The question is frozen, while its gold answer is determined by executing the program against professional financial data at evaluation time, and a daily job re-executes it with no human or language model in the loop. A memorized answer is thus likely outdated, and a model must reach the current one through tools and analysis. Construction is equally automatic: a language model only verbalizes programs enumerated from real tool responses and never chooses an answer. Under general web search, six contemporary tool-calling models reach 70.9% to 76.7% answer accuracy. Over three consecutive days, 70% of served questions change their gold value, and even under tolerance-aware matching, a literally reused earlier snapshot fails on a majority of questions within a single day. Per-question tracking uncovers dynamic shifts in correctness masked by aggregate accuracy.
Long-horizon agents performing multi-step Web tasks must reuse earlier page observations and interactions to support subsequent actions. However, preserving relevant history does not ensure that it exposes the relations needed for the current decision. We call this gap representation mismatch. To this end, we propose MemPrism, a task-conditioned relational memory framework that separates persistent experience storage from decision-time working memory. MemPrism records interactions as the event stream and dynamically constructs relational views according to the current task context. A lightweight view policy selects the relation structure, evidence range, outcome condition, and granularity, while a deterministic composer and renderer transform historical facts into a temporary optical working-memory view for a frozen task policy. Experiments on long-horizon embodied and web-agent benchmarks show that MemPrism consistently improves the task performance, especially as trajectories become longer, while reducing memory token consumption. Furthermore, the learned view policy transfers across different VLMs without additional adaptation, demonstrating the effectiveness of task-conditioned relational views as a general memory interface for agents.
Haoxi Li, Sikai Bai, Jianfei Ma, Qinglin Hou, Xiaocheng Lu, Tao Han, Jie Zhang, Song Guo
Learning from experience is essential for building the next generation of agents that continually expand their capabilities. Pretrained vision-language models (VLMs) already internalize rich world knowledge, yet this knowledge is rarely leveraged to drive policy improvement through online interaction. Realizing this potential efficiently remains challenging, as policy updates shift the agent's interaction distribution, requiring the world model to adapt accordingly, while continually updating large pretrained models is computationally expensive. To address this challenge, we introduce Dwell, a model-based reinforcement learning (RL) framework that reframes a VLM as a recurrent latent transition operator, endowing it with the ability to predict, adjust, and act in latent space. Leveraging the VLM's own representations and its dual role as both world model and policy, the agent continually learns from real interactions to anticipate and evaluate action outcomes, and from latent imagination to improve its behavior. Across five interactive visual environments, Dwell achieves the strongest overall performance among trainable methods while substantially reducing the compute required for imagination, showing that effective world models for VLM agents need not reconstruct the future; they need to predict futures that are useful for action. Our code is publicly available at https://github.com/Tom111-lhx/Dwell
Xing Liu, Yifan Guo, Hongtai Lv, Lei Lou, Yuyi Tan, Jiang Du, Xijia Zhang, Ming Zhang 等 21 位作者
Intraoperative frozen-section diagnosis of central nervous system (CNS) lesions guides surgical decisions but is challenged by artefacts, diagnostic complexity and time constraints. Most pathology foundation models have limited exposure to CNS frozen tissue. Here we present FOCUS, a subspecialty foundation model pretrained on 27,930 frozen-section whole-slide images enriched for CNS tissue. FOCUS ranked first across all 11 diagnostic and 12 biomarker-prediction tasks in the internal test cohort, achieving mean AUCs of 0.969 and 0.877, respectively. This advantage was retained in the temporal and external test cohorts. We further organized a WHO-aligned diagnostic hierarchy, where FOCUS also outperformed comparator models. In a reader study, FOCUS assistance enabled junior readers to approach or exceed unassisted senior readers in hierarchical diagnostic accuracy and granularity. More importantly, FOCUS assistance enabled readers to correctly identify a mean of 63.0% of initially missed lymphoma cases (73.2% among junior readers), potentially avoiding unnecessary extensive resection. These findings highlight the potential of FOCUS to provide clinically actionable support for intraoperative CNS pathology and inform surgical decisions.
Xu Zheng, Wei Cheng, Zhuomin Chen, Dongkuan Xu, Mo Sha, Jingchao Ni, Dongsheng Luo
On-device fine-tuning lets a large language model (LLM) learn from its user's private data without uploading it. Low-rank adaptation (LoRA) reduces the trainable parameters, but each first-order training step still multiplies by the frozen base-model weights during the forward pass, activation recomputation, and input-gradient computation. A phone's neural processing unit (NPU) already holds these weights for inference; using it for training also requires computing input gradients, which its inference interface does not support. We present a system that runs all three frozen-weight products on the NPU while keeping the training graph, adapters, and optimizer on the CPU. Its input-gradient kernels reuse the resident weight tiles and the forward dequantizer, avoiding a transposed weight copy and ensuring forward and backward use the same effective weights. To preserve usable gradients, the system selects precision by gradient sensitivity and applies a per-row range guard for half-precision (FP16) outputs. Its runtime fuses the remaining CPU operators and overlaps NPU execution with adapter computation, while resumable execution and thermal pacing let training proceed as a background job. We evaluate it on two Qualcomm Snapdragon platforms using four LLMs from the Qwen and Llama families. Our workloads cover nightly personalization with labels supplied on the phone and the public LaMP-2 benchmark; comparisons include an optimized CPU trainer, a phone-GPU trainer, forward-only training, and GPU-trained references.
Geometric alignment scores based on Gram determinants provide a compact way to model higher-order consistency among modalities, yet how such scores respond to modality degradation is poorly understood. This paper asks whether the response of a multimodal geometric score is determined primarily by the magnitude of the perturbation-induced displacement. Using frozen cohorts from MSR-VTT (N=878) and DiDeMo (N=980), we apply controlled video blur and audio noise and analyze the response in the relational geometry on which the score is defined. Displacement magnitude explains at most 15% of the out-of-sample variance in the absolute response, and magnitude-matched pairs respond systematically differently, so scalar magnitude does not organize the response. The closed-form first-order expansion of the Gramian volume yields the Directional Geometric Response (DGR): the projection of the displacement onto the local volume gradient, which jointly captures the clean operating point, displacement magnitude, and displacement direction. The absolute first-order DGR term explains the observed response with out-of-sample R^2 of 0.838-0.969, matched-magnitude ranking accuracies of 0.864-0.963, and response-sign accuracies of 0.909-0.989, whereas the tested direction-free alternatives remain weak or unstable under the corresponding evaluation protocols. A pre-specified gain-normalization candidate, V/(g_V+eps), fails its predictability and clean-order gates. DGR uses the observed degraded-state displacement and is therefore an explanatory quantity, not a deployment-time predictor: geometric response depends on where the representation operates, how far degradation moves the relational geometry, and in which direction it moves.
Evaluating human-AI co-design requires testing whether coding agents can translate real users' multimodal requests into working application revisions. We introduce Co-Design Bench, a reference-free benchmark of 273 tasks curated from 37,463 deduplicated queries across 26,314 real user coding sessions. Its two tracks, Multimodal Editing (ME) and Repository Editing (RE), cover localized edits with optional image, audio, video, or document inputs and coordinated repository revisions. Tasks span three capability dimensions: Grounded Understanding, UI Transformation, and Functional Implementation. The tasks pair normalized requests and runnable pre-edit applications with 2,331 hidden expert-authored acceptance criteria. Our judge-then-refute protocol grades criteria using source and browser evidence, challenges positive verdicts with a separate LLM verifier, and audits visual and interaction claims without gold code or target screenshots. Under a shared agentic harness, the strongest of ten model endpoints passes an average of 69.9% of criteria per task but fully satisfies only 26.4% of tasks. All ten endpoints achieve lower criterion completion on repository edits spanning three or four change types than on those spanning one or two. Co-Design Bench provides a testbed for studying the gap between partial requirement satisfaction and complete task success under agentic execution for real-user application revisions.
Han Zhang, Zihan Gu Institute of Information Engineering, Chinese Academy of Sciences, Zhiyuan Wang, Tianyi Ma, Jiacheng Lu Shanghai Jiao Tong University, Xinyan Zhang independent, Yuhao Wei, Cheng Hua Shanghai Jiaotong University
While Large Language Models (LLMs) achieve high accuracy on established Classical Chinese Poetry benchmarks, it remains challenging to distinguish transferable Linguistic-Aesthetic Reasoning from reliance on familiar pre-training patterns. To address this issue, we introduce Neo-Classic, an evaluation benchmark that combines a constructionist Out-of-Sample (OOS) dataset with a suite of reverse understanding probes. Unlike traditional benchmarks that rely on verification or generation over historical corpora, Neo-Classic comprises strictly metrical poetry authored by contemporary experts, reducing the possibility of direct retrieval. We evaluate state-of-the-art models, including Qwen3-Max, Gemini-3-Pro, and DeepSeek-V3.2, across five behavioral probes designed to test hierarchical constraint satisfaction. Our results reveal two primary limitations. First, a performance gap of 20%–50% emerges when models transition from historical to contemporary texts. Second, models exhibit substantial difficulties in discourse-level ordering tasks, with standard accuracy remaining low (0–13%). Although expert-level guidance improves the performance of reasoning-enhanced models to 36%, a notable gap with human experts persists. These findings suggest that while current LLMs capture local formal patterns, they struggle with global hierarchical planning required for robust Linguistic-Aesthetic 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
Retrieval-Augmented Generation (RAG) enhances the response capabilities of language models by integrating external knowledge sources. However, document chunking as an important part of RAG system often lacks effective evaluation tools. This paper first analyzes why existing RAG evaluation benchmarks are inadequate for assessing document chunking quality, specifically due to evidence sparsity. Based on this conclusion, we propose HiCBench, which includes manually annotated multi-level document chunking points, synthesized evidence-dense question answer(QA) pairs, and their corresponding evidence sources. We also propose HiChunk, a hierarchical document structuring framework using fine-tuned LLMs and the Auto-Merge retrieval algorithm to enhance retrieval quality. Experiments demonstrate that HiCBench effectively evaluates the impact of different chunking methods across the entire RAG pipeline. Moreover, HiChunk achieves better chunking quality within reasonable time consumption, thereby enhancing the overall performance of RAG systems. Source code is available at https://github.com/TencentCloudADP/hichunk.
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
Zihan Gu Institute of Information Engineering, Chinese Academy of Sciences, TianYi Zhang University of the Chinese Academy of Sciences, Xinyan Zhang independent, Zhiyuan Wang, Han Zhang, Yuhao Wei, Jiacheng Lu Shanghai Jiao Tong University, Tianyi Ma 等 11 位作者
Model editing provides a promising mechanism for updating large language models (LLMs) without expensive retraining. Existing approaches, particularly locate-and-edit methods based on least-squares optimization, aim to introduce targeted knowledge changes while preserving pre-trained behavior. In this work, we show that this objective is fundamentally fragile under standard single-edit evaluation protocols. We first develop a unified theoretical framework that characterizes activation-based editing as a constrained intervention on intermediate representations. Within this framework, we demonstrate that least-squares edits cannot, in general, isolate target updates from unrelated activations, giving rise to unavoidable interference that accumulates with successive edits. Crucially, this degradation can remain undetected in single-edit settings when assessed using conventional success and locality metrics. To expose such hidden instabilities, we introduce an uncertainty-based evaluation protocol that combines structured semantic perturbations with uncertainty quantification based on Sampling with Perturbation for UQ. By measuring edit-induced growth in aleatoric and epistemic uncertainty, our method reveals local knowledge conflicts that are invisible to existing benchmarks. Extensive experiments across multiple models, datasets, and editing algorithms show that both least-squares and other parameter-update-based methods consistently increase post-edit uncertainty. Together, our results suggest that current evaluation practices substantially overestimate the reliability of single-edit model editing, and that uncertainty-based diagnostics are necessary for assessing edit stability.
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
Zheng Luo, T Pranav Kutralingam, Ogochukwu N. Okoani, Wanpeng Xu Arizona State University, Hua Wei Arizona State University, Xiyang Hu Arizona State University
Large Language Models (LLMs) are increasingly deployed as agents that invoke external tools through structured function calls. While recent work reports strong tool-calling performance under standard English-centric evaluations, the robustness of tool calling under multilingual user interactions remains underexplored. In this work, we introduce MLCL, a diagnostic benchmark, and conduct a systematic evaluation of multilingual tool calling across Chinese, Hindi, and the low-resource language Igbo. Through fine-grained error analysis, we show that many failures occur despite correct intent understanding and tool selection. We identify parameter value language mismatch as a dominant failure mode, where models generate semantically appropriate parameter values in the user’s language, violating language-invariant execution conventions. We further evaluate several inference-time system strategies and find that while these strategies substantially reduce language-induced execution errors, none of them can fully recover English-level performance.
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
Yandi Wang, Libin Zhan Zhejiang University, Ziwei Huang, Tiancheng Luo, Yuxuan Jiang, Wang Dong Zhejiang University, Leilei Gan Zhejiang University, Jun Chen Zhejiang University
Extracting structured information from visual documents (Visual Information Extraction, VIE) is a cornerstone of business automation. While recent Multimodal Large Language Models (MLLMs) have shown promising capabilities, existing benchmarks suffer from critical limitations in scale and realism, lack semantic granularity, and fail to cover diverse document types. To bridge this gap, we introduce ReceiptBench, a large-scale, human-annotated benchmark consisting of 10k diverse receipts, organizing information extraction into four hierarchical sub-tasks: (1) Basic Perception for raw text spotting, (2) Format Normalization for strictly following standardization instructions, (3) Semantic Reasoning for inferring implicit attributes from context, and (4) Structure Parsing for handling nested line items. Furthermore, we propose a two-stage training framework incorporating Metric-Aware Group Relative Policy Optimization (GRPO), which translates rigorous evaluation constraints into reinforcement learning signals to enhance structural consistency. Extensive experiments demonstrate that our method yields state-of-the-art performance, surpassing leading proprietary models on complex reasoning tasks. We release our datasets and code at https://github.com/wwwT0ri/ReceiptBench.
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
Kangcheng Luo Peking University, Tinglang Wu, Yansong Feng Peking University
Recent search-augmented LLMs trained with reinforcement learning (RL) can interleave searching and reasoning for multi-hop reasoning tasks. However, they face two critical failure modes as the accumulating context becomes flooded with both crucial evidence and irrelevant information: (1) ineffective search chain construction that produces incorrect queries or omits retrieval of critical information, and (2) reasoning hijacking by peripheral evidence that causes models to misidentify distractors as valid evidence.To address these challenges, we propose D²Plan, a Dual-agent Dynamic global Planning paradigm for complex retrieval-augmented reasoning. D²Plan operates through the collaboration of a Reasoner and a Purifier: the Reasoner constructs explicit global plans during reasoning and dynamically adapts them based on retrieval feedback; the Purifier assesses retrieval relevance and condenses key information for the Reasoner.We further introduce a two-stage training framework consisting of supervised fine-tuning (SFT) cold-start on synthesized trajectories and RL with plan-oriented rewards to teach LLMs to master the D²Plan paradigm. Extensive experiments demonstrate that D²Plan enables more coherent multi-step reasoning and stronger resilience to irrelevant information, thereby achieving superior performance on challenging QA benchmarks.
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
Zhensheng Luo Zhejiang University, Sai Wu Zhejiang University, Yuan Qiu, Chang Yao Zhejiang University, Gang Chen College of Computer Science and Technology, Zhejiang University, Xiu Tang Zhejiang University
Natural language to SQL (NL2SQL) provides an intuitive interface for querying structured data, yet real user questions are often noisy, ambiguous, and weakly grounded to database semantics.As a result, token-level schema linking and single-pass SQL decoding can be brittle: small misunderstandings in language or schema grounding may propagate into incorrect generation.We present QBridge, an agentic, feedback-driven NL2SQL framework based on a Refined Gold Query Paradigm, which bridges natural language and SQL via Gold Query—a structured, SQL-aligned intermediate representation.A core insight of QBridge is Distilled Back-Translation (DBT) for SL-independent rewriting.DBT converts SQL-grounded supervision into execution-verified Gold-Query-style rewrites from a teacher model, and distills a lightweight, plug-and-play rewriter that generates schema-aware rewrites without requiring explicit schema linking at inference.QBridge then (i) verifies and conservatively refines the rewrite into a high-fidelity Refined Gold Query, and (ii) refines the generated SQL with dual feedback from execution validity and semantic consistency, enabling interpretable self-correction while remaining compatible with diverse SQL backbones.Extensive experiments on Spider, BIRD, and three robustness variants demonstrate that QBridge consistently improves zero-shot NL2SQL, outperforming strong prompting and agentic baselines while showing strong robustness and generalization. Code and data are available at https://github.com/WannaBSteve/QBridge.
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
Evaluating expressive speech remains challenging, as existing methods mainly assess emotional intensity and overlook whether a speech sample is expressively appropriate for its contextual setting. This limitation hinders reliable evaluation of speech systems used in narrative-driven and interactive applications, such as audiobooks and conversational agents. We introduce CEAEval, a Context-rich framework for Evaluating Expressive Appropriateness in speech, which assesses whether a speech sample expressively aligns with the underlying communicative intent implied by its discourse-level narrative context. To support this task, we construct CEAEval-D, the first context-rich speech dataset with real human performances in Mandarin conversational speech, providing narrative descriptions together with fifteen dimensions of human annotations covering expressive attributes and expressive appropriateness. We further develop CEAEval-M, a model that integrates knowledge distillation, planner-based multi-model collaboration, adaptive audio attention bias, and reinforcement learning to perform context-rich expressive appropriateness evaluation. Experiments on a human-annotated test set demonstrate that CEAEval-M substantially outperforms existing speech evaluation and analysis systems.
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
Chaoren Wang, Heng Lu, Xueyao Zhang The Chinese University of Hong Kong, Shenzhen, Shujie Liu, Yan Lu Microsoft Research Asia, Jinyu Li Microsoft, Zhizheng Wu Chinese University of Hong Kong, Shenzhen
Although Speech Large Language Models have achieved notable progress, a substantial modality reasoning gap remains: their reasoning performance on speech inputs is markedly weaker than on text. This gap could be associated with representational drift across Transformer layers and behavior deviations in long-chain reasoning. To address this issue, we introduce TARS, a reinforcement-learning framework that aligns text-conditioned and speech-conditioned trajectories through an asymmetric reward design. The framework employs two dense and complementary signals: representation alignment, which measures layer-wise hidden-state similarity between speech- and text-conditioned trajectories, and behavior alignment, which evaluates semantic consistency between generated outputs and reference text completions. Experiments on challenging reasoning benchmarks, including MMSU and OBQA, show that our approach significantly narrows the modality reasoning gap and achieves state-of-the-art performance among 7B-scale Speech LLMs.
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