Weijie Shi

Chi-Min Chan, Jianxuan Yu, Weize Chen et al. · tsinghua

The rapid advancement of large language models (LLMs) has led to the rise of LLM-based agents. Recent research shows that multi-agent systems (MAS), where each agent plays a specific role, can outperform individual LLMs. However, configuring an MAS for a task remains challenging, with performance only observable post-execution. Inspired by scaling laws in LLM development, we investigate whether MAS performance can be predicted beforehand. We introduce AgentMonitor, a framework that integrates at the agent level to capture inputs and outputs, transforming them into statistics for training a regression model to predict task performance. Additionally, it can further apply real-time corrections to address security risks posed by malicious agents, mitigating negative impacts and enhancing MAS security. Experiments demonstrate that an XGBoost model achieves a Spearman correlation of 0.89 in-domain and 0.58 in more challenging scenarios. Furthermore, using AgentMonitor reduces harmful content by 6.2% and increases helpful content by 1.8% on average, enhancing safety and reliability. Code is available at \url{https://github.com/chanchimin/AgentMonitor}.

Hao Chen, Xing Tang, Qirui Liu et al.

Large Language Models (LLMs) have significantly advanced online data services, particularly in the domain of financial question answering (FinQA). However, such systems remain susceptible to numerical reasoning hallucinations, which critically undermine reliability in high-stakes financial applications. Although retrieval-augmented generation (RAG) has been widely adopted to ground responses in external knowledge, it introduces three persistent challenges: noise sensitivity, calculation fragility, and an auditability crisis. Existing model-centric approaches, which primarily focus on optimizing either the retriever or generator in isolation, still struggle to address these issues in an integrated manner. In this work, we pioneer a data-centric paradigm and propose a novel framework, the Data-centric Reasoning Compiler (DCRC). The framework operates through three cohesive phases: (1) adversarial data construction, which synthesizes training examples with controlled noise to teach robustness; (2) multi-stage training that cultivates a Data-centric Structuring Agent (DSA) capable of explicit evidence auditing and program synthesis; and (3) a compile-and-execute inference process, where the DSA transforms user queries and retrieved documents into verifiable, executable reasoning programs. This data-driven framework ensures faithful numerical reasoning by design. We conduct extensive experiments on established offline benchmarks and further validate our framework through deployment in a real-world online financial QA system.

Jiawei Li, Ziyi Liu, Weijie Shi et al.

3D object grounding localizes referred objects in a 3D scene from natural language. Unified instance-centric 3D-LLMs aim to solve grounding together with dialog, QA, and captioning, yet many rely on a single pointer-style grounding decision that compresses a relational instruction into one selection. This is brittle for fine-grained queries where multiple same-class candidates must be ruled out by context objects and spatial relations. We propose Structured Spatial Reasoning 3D-LLM (SSR3D-LLM), a structured grounding interface for unified 3D-LLMs. Given fixed Mask3D object proposals, the LLM writes a sequence of latent spatial reasoning steps and memory tokens from the query, and a geometry-aware scorer reads these latent steps in order to refine candidate rankings step by step with step-length masking. The latent steps are learned from standard benchmark target supervision with auxiliary referential-cue supervision during training, while inference uses only the input query and Mask3D proposals. Across ReferIt3D, ScanRefer, and Multi3DRef, SSR3D-LLM achieves the strongest results among unified 3D-LLM baselines, with substantial gains over the single-pointer QPG baseline on fine-grained grounding and consistent improvements over prior unified 3D-LLMs, while preserving the default language-task route.

Weijie Shi, Yanxi Chen, Zexi Li et al.

Reinforcement learning drives recent advances in LLM reasoning and agentic capabilities, yet current approaches struggle with both exploration and exploitation. Exploration suffers from low success rates on difficult tasks and high costs of repeated rollouts from scratch. Exploitation suffers from coarse credit assignment and training instability: Trajectory-level rewards penalize valid prefixes for later errors, and failure-dominated groups overwhelm the few positive signals, leaving optimization without constructive direction. To this end, we propose R$^3$L, Reflect-then-Retry Reinforcement Learning with Language-Guided Exploration, Pivotal Credit, and Positive Amplification. To synthesize high-quality trajectories, R$^3$L shifts from stochastic sampling to active synthesis via reflect-then-retry, leveraging language feedback to diagnose errors, transform failed attempts into successful ones, and reduce rollout costs by restarting from identified failure points. With errors diagnosed and localized, Pivotal Credit Assignment updates only the diverging suffix where contrastive signals exist, excluding the shared prefix from gradient update. Since failures dominate on difficult tasks and reflect-then-retry produces off-policy data, risking training instability, Positive Amplification upweights successful trajectories to ensure positive signals guide the optimization process. Experiments on agentic and reasoning tasks demonstrate 5\% to 52\% relative improvements over baselines while maintaining training stability. Our code is released at https://github.com/shiweijiezero/R3L.

Qirui Liu, Hao Chen, Weijie Shi et al.

Accurately identifying student misconceptions is crucial for personalized education but faces three challenges: (1) data scarcity with long-tail distribution, where authentic student reasoning is difficult to synthesize; (2) fuzzy boundaries between error categories with high annotation noise; (3) deployment parado-large models overlook unconventional approaches due to pretraining bias and cannot be deployed on edge, while small models overfit to noise. Unlike traditional methods that increase diversity through large-scale data synthesis, we propose a two-stage knowledge distillation framework that mines high-value samples from existing data. The first stage performs standard distillation to transfer task capabilities. The second stage introduces a dual-layer marginal selection mechanism based on cognitive uncertainty, identifying four types of critical samples based on teacher model uncertainty and confidence differences. For different data subsets, we design difficulty-adaptive mechanism to balance hard/soft label contributions, enabling student models to inherit inter-class relationships from teacher soft labels while distinguishing ambiguous error types. Experiments show that with augmented training on only 10.30% of filtered samples, we achieve MAP@3 of 0.9585 (+17.8%) on the MAP-Charting dataset, and using only a 4B parameter model, we attain 84.38% accuracy on cross-topic tests of middle school algebra misconception benchmarks, significantly outperforming sota LLM (67.73%) and standard fine-tuned 72B models (81.25%). Our code is available at https://github.com/RoschildRui/acl2026_map.

Dongjie Xu, Hao Wu, Weijie Shi et al.

Through systematic experiments on long-context generation, we observe a damaging failure mode in which decoding can collapse into persistent repetition loops. We find that this degeneration is driven by collapsed attention patterns, where a subset of heads locks onto a narrow suffix of the history, and is further stabilized by inference-time KV cache reuse. Crucially, since many existing KV cache policies rely on attention-based importance, this collapse can produce spuriously high scores for repetitive tokens, causing cache management to inadvertently amplify repetition. To study this phenomenon in a controlled and reproducible manner, we introduce LoopBench, a benchmark with explicit loop-inducing conditions and loop-oriented metrics that quantify repetition severity and generation instability beyond downstream task scores. Building on these insights, we propose LoopGuard, a lightweight, plug-in KV cache guard that detects loop onset online and disrupts the feedback cycle by pruning repetitive tail spans under a fixed cache budget. Experiments on LoopBench show that LoopGuard reduces loop incidence by over 90 percentage points, while restoring output diversity and reducing token waste.

Xing Tang, Hao Chen, Shiwei Li et al.

Large language models (LLMs) have been incorporated into numerous industrial applications. Meanwhile, a vast array of API assets is scattered across various functions in the financial domain. An online financial question-answering system can leverage both LLMs and private APIs to provide timely financial analysis and information. The key is equipping the LLM model with function calling capability tailored to a financial scenario. However, a generic LLM requires customized financial APIs to call and struggles to adapt to the financial domain. Additionally, online user queries are diverse and contain out-of-distribution parameters compared with the required function input parameters, which makes it more difficult for a generic LLM to serve online users. In this paper, we propose a data-driven pipeline to enhance function calling in LLM for our online, deployed financial QA, comprising dataset construction, data augmentation, and model training. Specifically, we construct a dataset based on a previous study and update it periodically, incorporating queries and an augmentation method named AugFC. The addition of user query-related samples will \textit{exploit} our financial toolset in a data-driven manner, and AugFC explores the possible parameter values to enhance the diversity of our updated dataset. Then, we train an LLM with a two-step method, which enables the use of our financial functions. Extensive experiments on existing offline datasets, as well as the deployment of an online scenario, illustrate the superiority of our pipeline. The related pipeline has been adopted in the financial QA of YuanBao\footnote{https://yuanbao.tencent.com/chat/}, one of the largest chat platforms in China.

Kai Qian, Weijie Shi, Jiaqi Wang et al.

Multimodal fake news detection (MFND) aims to verify news credibility by jointly exploiting textual and visual evidence. However, real-world news dissemination frequently suffers from missing modality due to deleted images, corrupted screenshots, and similar issues. Thus, robust detection in this scenario requires preserving strong verification ability for each modality, which is challenging in MFND due to insufficient learning of the low-contribution modality and scarce unimodal annotations. To address this issue, we propose Head-wise Modality Specialization within Multimodal Large Language Models (MLLMs) for robust MFND under missing modality. Specifically, we first systematically study attention heads in MLLMs and their relationship with performance under missing modality, showing that modality-critical heads serve as key carriers of unimodal verification ability through their modality specialization. Based on this observation, to better preserve verification ability for the low-contribution modality, we introduce a head-wise specialization mechanism that explicitly allocates these heads to different modalities and preserves their specialization through lower-bound attention constraints. Furthermore, to better exploit scarce unimodal annotations, we propose a Unimodal Knowledge Retention strategy that prevents these heads from drifting away from the unimodal knowledge learned from limited supervision. Experiments show that our method improves robustness under missing modality while preserving performance with full multimodal input.

Kwun Hang Lau, Ruiyuan Zhang, Weijie Shi et al.

While Retrieval-Augmented Generation (RAG) excels at injecting static, factual knowledge into Large Language Models (LLMs), it exhibits a critical deficit in handling longitudinal queries that require tracking entities and phenomena across time. This blind spot arises because conventional, semantically-driven retrieval methods are not equipped to gather evidence that is both topically relevant and temporally coherent for a specified duration. We address this challenge by proposing a new framework that fundamentally redesigns the RAG pipeline to infuse temporal logic. Our methodology begins by disentangling a user's query into its core subject and its temporal window. It then employs a specialized retriever that calibrates semantic matching against temporal relevance, ensuring the collection of a contiguous evidence set that spans the entire queried period. To enable rigorous evaluation of this capability, we also introduce the Analytical Diachronic Question Answering Benchmark (ADQAB), a challenging evaluation suite grounded in a hybrid corpus of real and synthetic financial news. Empirical results on ADQAB show that our approach yields substantial gains in answer accuracy, surpassing standard RAG implementations by 13% to 27%. This work provides a validated pathway toward RAG systems capable of performing the nuanced, evolutionary analysis required for complex, real-world questions. The dataset and code for this study are publicly available at https://github.com/kwunhang/TA-RAG.

Weijie Shi, Han Zhu, Jiaming Ji et al.

Legal judgment prediction (LJP) aims to function as a judge by making final rulings based on case claims and facts, which plays a vital role in the judicial domain for supporting court decision-making and improving judicial efficiency. However, existing methods often struggle with logical errors when conducting complex legal reasoning. We propose LegalReasoner, which enhances LJP reliability through step-wise verification and correction of the reasoning process. Specifically, it first identifies dispute points to decompose complex cases, and then conducts step-wise reasoning while employing a process verifier to validate each step's logic from correctness, progressiveness, and potential perspectives. When errors are detected, expert-designed attribution and resolution strategies are applied for correction. To fine-tune LegalReasoner, we release the LegalHK dataset, containing 58,130 Hong Kong court cases with detailed annotations of dispute points, step-by-step reasoning chains, and process verification labels. Experiments demonstrate that LegalReasoner significantly improves concordance with court decisions from 72.37 to 80.27 on LLAMA-3.1-70B. The data is available at https://huggingface.co/datasets/weijiezz/LegalHK.

Chuxue Cao, Zhenghao Zhu, Junqi Zhu et al.

Multilingual understanding is crucial for the cross-cultural applicability of Large Language Models (LLMs). However, evaluation benchmarks designed for Hong Kong's unique linguistic landscape, which combines Traditional Chinese script with Cantonese as the spoken form and its cultural context, remain underdeveloped. To address this gap, we introduce HKMMLU, a multi-task language understanding benchmark that evaluates Hong Kong's linguistic competence and socio-cultural knowledge. The HKMMLU includes 26,698 multi-choice questions across 66 subjects, organized into four categories: Science, Technology, Engineering, and Mathematics (STEM), Social Sciences, Humanities, and Other. To evaluate the multilingual understanding ability of LLMs, 90,550 Mandarin-Cantonese translation tasks were additionally included. We conduct comprehensive experiments on GPT-4o, Claude 3.7 Sonnet, and 18 open-source LLMs of varying sizes on HKMMLU. The results show that the best-performing model, DeepSeek-V3, struggles to achieve an accuracy of 75\%, significantly lower than that of MMLU and CMMLU. This performance gap highlights the need to improve LLMs' capabilities in Hong Kong-specific language and knowledge domains. Furthermore, we investigate how question language, model size, prompting strategies, and question and reasoning token lengths affect model performance. We anticipate that HKMMLU will significantly advance the development of LLMs in multilingual and cross-cultural contexts, thereby enabling broader and more impactful applications.

Weijie Shi, Jipeng Zhang, Yaguang Wu et al.

Large language models (LLMs) are commonly trained on multi-domain datasets, where domain sampling strategies significantly impact model performance due to varying domain importance across downstream tasks. Existing approaches for optimizing domain-level sampling strategies struggle with maintaining intra-domain consistency and accurately measuring domain impact. In this paper, we present Domain Impact-aware Data Sampling (DIDS). To ensure intra-domain consistency, a gradient clustering algorithm is proposed to group training data based on their learning effects, where a proxy language model and dimensionality reduction are employed to reduce computational overhead. To accurately measure domain impact, we develop a Fisher Information Matrix (FIM) guided metric that quantifies how domain-specific parameter updates affect the model's output distributions on downstream tasks, with theoretical guarantees. Furthermore, to determine optimal sampling ratios, DIDS combines both the FIM-guided domain impact assessment and loss learning trajectories that indicate domain-specific potential, while accounting for diminishing marginal returns. Extensive experiments demonstrate that DIDS achieves 3.4% higher average performance while maintaining comparable training efficiency. The code is available at https://github.com/shiweijiezero/DIDS.

Hanghui Guo, Weijie Shi, Mengze Li et al.

Short-video misinformation detection has attracted wide attention in the multi-modal domain, aiming to accurately identify the misinformation in the video format accompanied by the corresponding audio. Despite significant advancements, current models in this field, trained on particular domains (source domains), often exhibit unsatisfactory performance on unseen domains (target domains) due to domain gaps. To effectively realize such domain generalization on the short-video misinformation detection task, we propose deep insights into the characteristics of different domains: (1) The detection on various domains may mainly rely on different modalities (i.e., mainly focusing on videos or audios). To enhance domain generalization, it is crucial to achieve optimal model performance on all modalities simultaneously. (2) For some domains focusing on cross-modal joint fraud, a comprehensive analysis relying on cross-modal fusion is necessary. However, domain biases located in each modality (especially in each frame of videos) will be accumulated in this fusion process, which may seriously damage the final identification of misinformation. To address these issues, we propose a new DOmain generalization model via ConsisTency and invariance learning for shORt-video misinformation detection (named DOCTOR), which contains two characteristic modules: (1) We involve the cross-modal feature interpolation to map multiple modalities into a shared space and the interpolation distillation to synchronize multi-modal learning; (2) We design the diffusion model to add noise to retain core features of multi modal and enhance domain invariant features through cross-modal guided denoising. Extensive experiments demonstrate the effectiveness of our proposed DOCTOR model. Our code is public available at https://github.com/ghh1125/DOCTOR.

Weijie Shi, Qiang Xu, Fan Deng et al.

Speculative decoding accelerates LLM inference by drafting a tree of candidate continuations and verifying it in one target forward. Existing drafters fall into two camps with opposite weaknesses. Autoregressive drafters such as EAGLE-3 preserve dependence along each draft path but call the drafter once per tree depth, making drafting a non-trivial share of per-iteration latency. Parallel drafters cut drafter calls by predicting multiple future positions in one forward, but each position is predicted without seeing the others, producing paths the verifier rejects. In this paper, we propose SpecBlock, a block-iterative drafter that combines path dependence with cheap drafting. Each drafter forward produces K dependent positions and we call this a block. The draft tree grows through repeated block expansions. Two mechanisms explicitly carry path dependence to keep later draft positions accurate. Within each block, a layer-wise shift carries the previous position's hidden state into every decoder layer. Across blocks, each new block can start from any position of the previous block, inheriting its hidden state to extend the path. To spend verifier budget where acceptance is likely, a co-trained rank head replaces the fixed top-k tree by allocating per-position branching during drafting. To avoid training the drafter on prefixes it never produces at inference, a valid-prefix mask drops the loss at later positions once an earlier one is wrong. Beyond static drafting, a cost-aware bandit at deployment uses free verifier feedback to update the drafter selectively, only when the expected throughput gain exceeds the update cost. Experiments show that SpecBlock improves mean speedup by 8-13% over EAGLE-3 at 44-52% of its drafting cost, and cost-aware adaptation extends this lead to 11-19%.

Jiawei Shen, Jia Zhu, Hanghui Guo et al.

Large Language Models (LLMs) have shown remarkable performance in multi-turn dialogue. However, in multi-turn dialogue, models still struggle to stay aligned with what has been established earlier, follow dependencies across many turns, and avoid drifting into incorrect facts as the interaction grows longer. Existing approaches primarily focus on extending the context window, introducing external memory, or applying context compression, yet these methods still face limitations such as \textbf{contextual inertia} and \textbf{state drift}. To address these challenges, we propose the \textbf{A}daptive \textbf{C}ontext \textbf{R}efactoring \textbf{(ACR)} Framework, which dynamically monitors and reshapes the interaction history to mitigate contextual inertia and state drift actively. ACR is built on a library of context refactoring operators and a teacher-guided self-evolving training paradigm that learns when to intervene and how to refactor, thereby decoupling context management from the reasoning process. Extensive experiments on multi-turn dialogue demonstrate that our method significantly outperforms existing baselines while reducing token consumption.

Jiaqi Wang, Wenhao Zhang, Weijie Shi et al.

On-policy distillation (OPD) has shown strong potential for transferring reasoning ability from frontier or domain-specific models to smaller students. While effective on static single-turn tasks, its behavior in multi-turn agent settings remains underexplored. In this work, we identify a key limitation of vanilla OPD in such settings, which we term Trajectory-Level KL Instability. Specifically, we observe that KL divergence increases together with a drop in success rate, and even after convergence, the KL remains high, leading to unstable training. This instability arises from inter-turn error compounding: as errors accumulate, the student is driven beyond the teacher's effective support, rendering the supervision signal unreliable. To address this, we propose TCOD (Temporal Curriculum On-Policy Distillation), a simple yet effective framework that controls the trajectory depth exposed to the student and progressively expands it from short to long with a curriculum schedule.Experimental results across four student-teacher pairs on three multi-turn agent benchmarks (ALFWorld, WebShop, ScienceWorld) show that TCOD mitigates KL escalation and enhances KL stability throughout training, improving agent performance by up to 18 points over vanilla OPD. Further evaluations show that TCOD can even surpass the teacher's performance and generalize to tasks on which the teacher fails.

Xuchen Pan, Yanxi Chen, Yushuo Chen et al.

Trinity-RFT is a general-purpose, unified and easy-to-use framework designed for reinforcement fine-tuning (RFT) of large language models. It is built with a modular and decoupled design, consisting of (1) an RFT-core that unifies and generalizes synchronous/asynchronous, on-policy/off-policy, and online/offline modes of RFT; (2) seamless integration for agent-environment interaction with high efficiency and robustness; and (3) systematic data pipelines optimized for RFT. Trinity-RFT can be easily adapted for diverse application scenarios, and serves as a unified platform for development and research of advanced reinforcement learning paradigms at both macroscopic and microscopic levels. This technical report outlines the vision, features, design and implementations of Trinity-RFT, accompanied by extensive examples, applications and experiments that demonstrate its functionalities and user-friendliness.

Yue Cui, Liuyi Yao, Shuchang Tao et al.

Large language models (LLMs) have significantly advanced natural language processing, particularly through the integration of external tools and APIs. However, their effectiveness is frequently hampered by parameter mis-filling during tool calling. In this paper, we propose the Hierarchical Tool Error Checklist (HiTEC) framework to systematically diagnose and mitigate tool-calling errors without relying on extensive real-world interactions. HiTEC introduces a two-tiered approach: a global error checklist that identifies common, cross-tool issues, and a local error checklist that targets tool-specific and contextual failures. Building on this structure, we propose two deployments: HiTEC-In Context Learning (HiTEC-ICL) and HiTEC-Kahneman-Tversky Optimization (HiTEC-KTO). HiTEC-ICL embeds the global checklist in the initial prompts and leverages a two-round conversational interaction to dynamically refine parameter handling, while HiTEC-KTO generates high-quality negative examples to drive fine-tuning via preference-based optimization. Extensive experiments across five public datasets demonstrate that our framework significantly improves parameter-filling accuracy and tool-calling success rates compared to baseline methods.

Weijie Shi, Chengyi Ju, Chengzhong Liu et al.

Do Large Language Models (LLMs) hold positions that conflict with your country's values? Occasionally they do! However, existing works primarily focus on ethical reviews, failing to capture the diversity of national values, which encompass broader policy, legal, and moral considerations. Furthermore, current benchmarks that rely on spectrum tests using manually designed questionnaires are not easily scalable. To address these limitations, we introduce NaVAB, a comprehensive benchmark to evaluate the alignment of LLMs with the values of five major nations: China, the United States, the United Kingdom, France, and Germany. NaVAB implements a national value extraction pipeline to efficiently construct value assessment datasets. Specifically, we propose a modeling procedure with instruction tagging to process raw data sources, a screening process to filter value-related topics and a generation process with a Conflict Reduction mechanism to filter non-conflicting values.We conduct extensive experiments on various LLMs across countries, and the results provide insights into assisting in the identification of misaligned scenarios. Moreover, we demonstrate that NaVAB can be combined with alignment techniques to effectively reduce value concerns by aligning LLMs' values with the target country.

Chuxue Cao, Mengze Li, Juntao Dai et al.

Large language models (LLMs) have shown promising first-order logic (FOL) reasoning capabilities with applications in various areas. However, their effectiveness in complex mathematical reasoning involving multi-step FOL deductions is still under-researched. While LLMs perform competitively on established mathematical reasoning benchmarks, they struggle with multi-step FOL tasks, as demonstrated by Deepseek-Prover-V2-7B's low accuracy (4.2%) on our proposed theorem proving dataset. This issue arises from the limited exploration of diverse proof strategies and the potential for early reasoning mistakes to undermine entire proofs. To address these issues, we propose DREAM, a self-adaptive solution that enhances the Diversity and REAsonability of LLMs' generation strategies. DREAM incorporates an Axiom-Driven Strategy Diversification mechanism to promote varied strategic outcomes and a Sub-Proposition Error Feedback to help LLMs reflect on and correct their proofs. Our contributions include pioneering advancements in LLMs' mathematical reasoning through FOL theorem proving, introducing a novel inference stage solution that improves performance by 0.6% to 6.4%, and providing a curated dataset of 447 mathematical theorems in Lean 4 format for evaluation.

Hanghui Guo, Jia Zhu, Shimin Di et al.

Dynamic Retrieval-augmented Generation (RAG) has shown great success in mitigating hallucinations in large language models (LLMs) during generation. However, existing dynamic RAG methods face significant limitations in two key aspects: 1) Lack of an effective mechanism to control retrieval triggers, and 2) Lack of effective scrutiny of retrieval content. To address these limitations, we propose an innovative dynamic RAG method, DioR (Adaptive Cognitive Detection and Contextual Retrieval Optimization), which consists of two main components: adaptive cognitive detection and contextual retrieval optimization, specifically designed to determine when retrieval is needed and what to retrieve for LLMs is useful. Experimental results demonstrate that DioR achieves superior performance on all tasks, demonstrating the effectiveness of our work.

Dongjie Xu, Yue Cui, Weijie Shi et al.

SQL query rewriting aims to reformulate a query into a more efficient form while preserving equivalence. Most existing methods rely on predefined rewrite rules. However, such rule-based approaches face fundamental limitations: (1) fixed rule sets generalize poorly to novel query patterns and struggle with complex queries; (2) a wide range of effective rewriting strategies cannot be fully captured by declarative rules. To overcome these issues, we propose using large language models (LLMs) to generate rewrites. LLMs can capture complex strategies, such as evaluation reordering and CTE rewriting. Despite this potential, directly applying LLMs often results in performance regressions or non-equivalent rewrites due to a lack of execution awareness and semantic grounding. To address these challenges, We present E3-Rewrite, an LLM-based SQL rewriting framework that produces executable, equivalent, and efficient queries. It integrates two core components: a context construction module and a reinforcement learning framework. First, the context module leverages execution plans and retrieved demonstrations to build bottleneck-aware prompts that guide inference-time rewriting. Second, we design a reward function targeting executability, equivalence, and efficiency, evaluated via syntax checks, equivalence verification, and cost estimation. Third, to ensure stable multi-objective learning, we adopt a staged curriculum that first emphasizes executability and equivalence, then gradually incorporates efficiency. Across multiple SQL benchmarks, our experiments demonstrate that E3-Rewrite can shorten query execution time by as much as 25.6% relative to leading baselines, while also producing up to 24.4% more rewrites that meet strict equivalence criteria. These gains extend to challenging query patterns that prior approaches could not effectively optimize.

Jiawei Shen, Jia Zhu, Hanghui Guo et al.

Temporal Knowledge Graph Reasoning (TKGR) aims to complete missing factual elements along the timeline. Depending on the temporal position of the query, the task is categorized into interpolation and extrapolation. Existing interpolation methods typically embed temporal information into individual facts to complete missing historical knowledge, while extrapolation techniques often leverage sequence models over graph snapshots to identify recurring patterns for future event prediction. These methods face two critical challenges: limited contextual modeling in interpolation and cognitive generalization bias in extrapolation. To address these, we propose a unified method for TKGR, dubbed DynaGen. For interpolation, DynaGen dynamically constructs entity-centric subgraphs and processes them with a synergistic dual-branch GNN encoder to capture evolving structural context. For extrapolation, it applies a conditional diffusion process, which forces the model to learn underlying evolutionary principles rather than just superficial patterns, enhancing its ability to predict unseen future events. Extensive experiments on six benchmark datasets show DynaGen achieves state-of-the-art performance. On average, compared to the second-best models, DynaGen improves the Mean Reciprocal Rank (MRR) score by 2.61 points for interpolation and 1.45 points for extrapolation.

Cui Yakun, Fushuo Huo, Weijie Shi et al.

The advent of multi-modal large language models (MLLMs) has greatly advanced research into applications for Video fake news detection (VFND) tasks. Traditional video-based FND benchmarks typically focus on the accuracy of the final decision, often failing to provide fine-grained assessments for the entire detection process, making the detection process a black box. Therefore, we introduce the MVFNDB (Multi-modal Video Fake News Detection Benchmark) based on the empirical analysis, which provides foundation for tasks definition. The benchmark comprises 10 tasks and is meticulously crafted to probe MLLMs' perception, understanding, and reasoning capacities during detection, featuring 9730 human-annotated video-related questions based on a carefully constructed taxonomy ability of VFND. To validate the impact of combining multiple features on the final results, we design a novel framework named MVFND-CoT, which incorporates both creator-added content and original shooting footage reasoning. Building upon the benchmark, we conduct an in-depth analysis of the deeper factors influencing accuracy, including video processing strategies and the alignment between video features and model capabilities. We believe this benchmark will lay a solid foundation for future evaluations and advancements of MLLMs in the domain of video fake news detection.

Guoqing Ma, Jia Zhu, Hanghui Guo et al.

Multi-agent systems (MAS) are critical for automating complex tasks, yet their practical deployment is severely hampered by the challenge of failure attribution. Current diagnostic tools, which rely on statistical correlations, are fundamentally inadequate; on challenging benchmarks like Who\&When, state-of-the-art methods achieve less than 15\% accuracy in locating the root-cause step of a failure. To address this critical gap, we introduce the first failure attribution framework for MAS grounded in multi-granularity causal inference. Our approach makes two key technical contributions: (1) a performance causal inversion principle, which correctly models performance dependencies by reversing the data flow in execution logs, combined with Shapley values to accurately assign agent-level blame; (2) a novel causal discovery algorithm, CDC-MAS, that robustly identifies critical failure steps by tackling the non-stationary nature of MAS interaction data. The framework's attribution results directly fuel an automated optimization loop, generating targeted suggestions whose efficacy is validated via counterfactual simulations. Evaluations on the Who\&When and TRAIL benchmarks demonstrate a significant leap in performance. Our method achieves up to 36.2\% step-level accuracy. Crucially, the generated optimizations boost overall task success rates by an average of 22.4\%. This work provides a principled and effective solution for debugging complex agent interactions, paving the way for more reliable and interpretable multi-agent systems.

Weijie Shi, Yue Cui, Hao Chen et al.

Cellular Trajectory Map-Matching (CTMM) aims to align cellular location sequences to road networks, which is a necessary preprocessing in location-based services on web platforms like Google Maps, including navigation and route optimization. Current approaches mainly rely on ID-based features and region-specific data to learn correlations between cell towers and roads, limiting their adaptability to unexplored areas. To enable high-accuracy CTMM without additional training in target regions, Zero-shot CTMM requires to extract not only region-adaptive features, but also sequential and location uncertainty to alleviate positioning errors in cellular data. In this paper, we propose a pixel-based trajectory calibration assistant for zero-shot CTMM, which takes advantage of transferable geospatial knowledge to calibrate pixelated trajectory, and then guide the path-finding process at the road network level. To enhance knowledge sharing across similar regions, a Gaussian mixture model is incorporated into VAE, enabling the identification of scenario-adaptive experts through soft clustering. To mitigate high positioning errors, a spatial-temporal awareness module is designed to capture sequential features and location uncertainty, thereby facilitating the inference of approximate user positions. Finally, a constrained path-finding algorithm is employed to reconstruct the road ID sequence, ensuring topological validity within the road network. This process is guided by the calibrated trajectory while optimizing for the shortest feasible path, thus minimizing unnecessary detours. Extensive experiments demonstrate that our model outperforms existing methods in zero-shot CTMM by 16.8\%.

Weijie Shi, Yue Cui, Yaguang Wu et al.

Diverse decoding of large language models is crucial for applications requiring multiple semantically distinct responses, yet existing methods primarily achieve lexical rather than semantic diversity. This limitation significantly constrains Best-of-N strategies, group-based reinforcement learning, and data synthesis. While temperature sampling and diverse beam search modify token distributions or apply n-gram penalties, they fail to ensure meaningful semantic differentiation. We introduce Semantic-guided Diverse Decoding (SemDiD), operating directly in embedding space that balances quality with diversity through three complementary mechanisms: orthogonal directional guidance, dynamic inter-group repulsion, and position-debiased probability assessment. SemDiD harmonizes these competing objectives using adaptive gain functions and constraint optimization, ensuring both quality thresholds and maximal semantic differentiation. Experiments show SemDiD consistently outperforms existing methods, improving Best-of-N coverage by 1.4-5.2% across diverse tasks and accelerating RLHF training convergence by 15% while increasing accuracy by up to 2.1%.

Yuanhong Zhang, Muyao Yuan, Weizhan Zhang et al.

The Segment Anything Model (SAM), a vision foundation model, exhibits impressive zero-shot capabilities in general tasks but struggles in specialized domains. Parameter-efficient fine-tuning (PEFT) is a promising approach to unleash the potential of SAM in novel scenarios. However, existing PEFT methods for SAM neglect the domain-invariant relations encoded in the pre-trained model. To bridge this gap, we propose InfoSAM, an information-theoretic approach that enhances SAM fine-tuning by distilling and preserving its pre-trained segmentation knowledge. Specifically, we formulate the knowledge transfer process as two novel mutual information-based objectives: (i) to compress the domain-invariant relation extracted from pre-trained SAM, excluding pseudo-invariant information as possible, and (ii) to maximize mutual information between the relational knowledge learned by the teacher (pre-trained SAM) and the student (fine-tuned model). The proposed InfoSAM establishes a robust distillation framework for PEFT of SAM. Extensive experiments across diverse benchmarks validate InfoSAM's effectiveness in improving SAM family's performance on real-world tasks, demonstrating its adaptability and superiority in handling specialized scenarios.

Yangping Chen, Weijie Shi, Mengze Li et al.

Multimodal fake news detection plays a crucial role in combating online misinformation. Unfortunately, effective detection methods rely on annotated labels and encounter significant performance degradation when domain shifts exist between training (source) and test (target) data. To address the problems, we propose ADOSE, an Active Domain Adaptation (ADA) framework for multimodal fake news detection which actively annotates a small subset of target samples to improve detection performance. To identify various deceptive patterns in cross-domain settings, we design multiple expert classifiers to learn dependencies across different modalities. These classifiers specifically target the distinct deception patterns exhibited in fake news, where two unimodal classifiers capture knowledge errors within individual modalities while one cross-modal classifier identifies semantic inconsistencies between text and images. To reduce annotation costs from the target domain, we propose a least-disagree uncertainty selector with a diversity calculator for selecting the most informative samples. The selector leverages prediction disagreement before and after perturbations by multiple classifiers as an indicator of uncertain samples, whose deceptive patterns deviate most from source domains. It further incorporates diversity scores derived from multi-view features to ensure the chosen samples achieve maximal coverage of target domain features. The extensive experiments on multiple datasets show that ADOSE outperforms existing ADA methods by 2.72\% $\sim$ 14.02\%, indicating the superiority of our model.

Yi Cao, Wei-Jie Xu, Yucheng Shen et al.

The rapid increase in the parameter counts of Large Language Models (LLMs), which often reach into the billions or even trillions, presents significant challenges for their practical deployment, particularly in resource-constrained environments. To address this issue, we propose PIP (Perturbation-based Iterative Pruning), a novel double-view structured pruning method to optimize LLMs, which combines information from two different views: the unperturbed view and the perturbed view. With the calculation of gradient differences, PIP iteratively prunes those that struggle to distinguish between these two views. Our experiments show that PIP reduces the parameter count by approximately 20% while retaining over 85% of the original model's accuracy across varied benchmarks. In some cases, the performance of the pruned model is within 5% of the unpruned version, demonstrating PIP's ability to preserve key aspects of model effectiveness. Moreover, PIP consistently outperforms existing state-of-the-art (SOTA) structured pruning methods, establishing it as a leading technique for optimizing LLMs in constrained environments.