Zhijiang Guo

Peilin Zhou, Zeqiang Wang, Dading Chong et al. · bytedance, harvard

The COVID-19 pandemic continues to bring up various topics discussed or debated on social media. In order to explore the impact of pandemics on people's lives, it is crucial to understand the public's concerns and attitudes towards pandemic-related entities (e.g., drugs, vaccines) on social media. However, models trained on existing named entity recognition (NER) or targeted sentiment analysis (TSA) datasets have limited ability to understand COVID-19-related social media texts because these datasets are not designed or annotated from a medical perspective. This paper releases METS-CoV, a dataset containing medical entities and targeted sentiments from COVID-19-related tweets. METS-CoV contains 10,000 tweets with 7 types of entities, including 4 medical entity types (Disease, Drug, Symptom, and Vaccine) and 3 general entity types (Person, Location, and Organization). To further investigate tweet users' attitudes toward specific entities, 4 types of entities (Person, Organization, Drug, and Vaccine) are selected and annotated with user sentiments, resulting in a targeted sentiment dataset with 9,101 entities (in 5,278 tweets). To the best of our knowledge, METS-CoV is the first dataset to collect medical entities and corresponding sentiments of COVID-19-related tweets. We benchmark the performance of classical machine learning models and state-of-the-art deep learning models on NER and TSA tasks with extensive experiments. Results show that the dataset has vast room for improvement for both NER and TSA tasks. METS-CoV is an important resource for developing better medical social media tools and facilitating computational social science research, especially in epidemiology. Our data, annotation guidelines, benchmark models, and source code are publicly available (https://github.com/YLab-Open/METS-CoV) to ensure reproducibility.

Siyang Song, Yuxin Song, Cheng Luo et al.

Graph is powerful for representing various types of real-world data. The topology (edges' presence) and edges' features of a graph decides the message passing mechanism among vertices within the graph. While most existing approaches only manually define a single-value edge to describe the connectivity or strength of association between a pair of vertices, task-specific and crucial relationship cues may be disregarded by such manually defined topology and single-value edge features. In this paper, we propose the first general graph representation learning framework (called GRATIS) which can generate a strong graph representation with a task-specific topology and task-specific multi-dimensional edge features from any arbitrary input. To learn each edge's presence and multi-dimensional feature, our framework takes both of the corresponding vertices pair and their global contextual information into consideration, enabling the generated graph representation to have a globally optimal message passing mechanism for different down-stream tasks. The principled investigation results achieved for various graph analysis tasks on 11 graph and non-graph datasets show that our GRATIS can not only largely enhance pre-defined graphs but also learns a strong graph representation for non-graph data, with clear performance improvements on all tasks. In particular, the learned topology and multi-dimensional edge features provide complementary task-related cues for graph analysis tasks. Our framework is effective, robust and flexible, and is a plug-and-play module that can be combined with different backbones and Graph Neural Networks (GNNs) to generate a task-specific graph representation from various graph and non-graph data. Our code is made publicly available at https://github.com/SSYSteve/Learning-Graph-Representation-with-Task-specific-Topology-and-Multi-dimensional-Edge-Features.

Zhicheng Yang, Yiwei Wang, Yinya Huang et al.

Large language models (LLMs) have exhibited their problem-solving abilities in mathematical reasoning. Solving realistic optimization (OPT) problems in application scenarios requires advanced and applied mathematics ability. However, current OPT benchmarks that merely solve linear programming are far from complex realistic situations. In this work, we propose OptiBench, a benchmark for End-to-end optimization problem-solving with human-readable inputs and outputs. OptiBench contains rich optimization problems, including linear and nonlinear programming with or without tabular data, which can comprehensively evaluate LLMs' solving ability. In our benchmark, LLMs are required to call a code solver to provide precise numerical answers. Furthermore, to alleviate the data scarcity for optimization problems, and to bridge the gap between open-source LLMs on a small scale (e.g., Llama-3-8b) and closed-source LLMs (e.g., GPT-4), we further propose a data synthesis method namely ReSocratic. Unlike general data synthesis methods that proceed from questions to answers, \ReSocratic first incrementally synthesizes formatted optimization demonstration with mathematical formulations step by step and then back-translates the generated demonstrations into questions. Based on this, we synthesize the ReSocratic-29k dataset. We further conduct supervised fine-tuning with ReSocratic-29k on multiple open-source models. Experimental results show that ReSocratic-29k significantly improves the performance of open-source models.

Xuming Hu, Zhijiang Guo, Guanyu Wu et al. · tsinghua

The explosion of misinformation spreading in the media ecosystem urges for automated fact-checking. While misinformation spans both geographic and linguistic boundaries, most work in the field has focused on English. Datasets and tools available in other languages, such as Chinese, are limited. In order to bridge this gap, we construct CHEF, the first CHinese Evidence-based Fact-checking dataset of 10K real-world claims. The dataset covers multiple domains, ranging from politics to public health, and provides annotated evidence retrieved from the Internet. Further, we develop established baselines and a novel approach that is able to model the evidence retrieval as a latent variable, allowing jointly training with the veracity prediction model in an end-to-end fashion. Extensive experiments show that CHEF will provide a challenging testbed for the development of fact-checking systems designed to retrieve and reason over non-English claims.

Jing Xiong, Jianhao Shen, Ye Yuan et al.

Automated theorem proving (ATP) has become an appealing domain for exploring the reasoning ability of the recent successful generative language models. However, current ATP benchmarks mainly focus on symbolic inference, but rarely involve the understanding of complex number combination reasoning. In this work, we propose TRIGO, an ATP benchmark that not only requires a model to reduce a trigonometric expression with step-by-step proofs but also evaluates a generative LM's reasoning ability on formulas and its capability to manipulate, group, and factor number terms. We gather trigonometric expressions and their reduced forms from the web, annotate the simplification process manually, and translate it into the Lean formal language system. We then automatically generate additional examples from the annotated samples to expand the dataset. Furthermore, we develop an automatic generator based on Lean-Gym to create dataset splits of varying difficulties and distributions in order to thoroughly analyze the model's generalization ability. Our extensive experiments show our proposed TRIGO poses a new challenge for advanced generative LM's including GPT-4 which is pre-trained on a considerable amount of open-source formal theorem-proving language data, and provide a new tool to study the generative LM's ability on both formal and mathematical reasoning.

Xiao Zhu, Xinyu Zhou, Boyu Zhu et al.

Reinforcement Learning from Verifiable Rewards (RLVR) has driven recent progress in code large language models by leveraging execution-based feedback from unit tests, but its scalability is fundamentally constrained by the availability and reliability of high-quality test cases. We propose CodeScaler, a reward model designed to scale both reinforcement learning training and test-time inference for code generation. CodeScaler is trained on carefully curated preference data derived from verified code problems and incorporates syntax-aware code extraction and validity-preserving reward shaping to ensure stable and robust optimization. Across four coding benchmarks, CodeScaler consistently outperforms execution-based RL by +1.55 points on Qwen3-8B-Base and +4.23 points on Qwen3-14B-Base. By further scaling to 44K problems with additional synthetic data, CodeScaler yields +14.64 points improvement over the base model without requiring any test cases. At inference time, CodeScaler serves as an effective test-time scaling method, achieving performance comparable to unit test approaches while providing a 10-fold reduction in latency. Moreover, CodeScaler surpasses existing reward models on RM-Bench not only in the code domain (+3.3 points), but also in general and reasoning domains (+2.7 points on average).

Zaoyu Chen, Jianbo Dai, Boyu Zhu et al.

Large language models (LLMs) can generate code from natural language, but the extent to which they capture intended program behavior remains unclear. Executable behavioral specifications, defined via preconditions and postconditions, provide a concrete means to assess such understanding. However, existing work on specification generation is constrained in evaluation methodology, task settings, and specification expressiveness. We introduce CodeSpecBench, a benchmark for executable behavioral specification generation under an execution-based evaluation protocol. CodeSpecBench supports both function-level and repository-level tasks and encodes specifications as executable Python functions. Constructed from diverse real-world codebases, it enables a realistic assessment of both correctness (accepting valid behaviors) and completeness (rejecting invalid behaviors). Evaluating 15 state-of-the-art LLMs on CodeSpecBench, we observe a sharp performance degradation on repository-level tasks, where the best model attains only a 20.2% pass rate. We further find that specification generation is substantially more challenging than code generation, indicating that strong coding performance does not necessarily reflect deep understanding of intended program semantics. Our data and code are available at https://github.com/SparksofAGI/CodeSpecBench.

Mengyi Deng, Zhiwei Li, Xin Li et al.

Large Language Model (LLM) agents often operate under underspecified user instructions, where latent uncertainty over user intent leads to erroneous tool actions. To address this challenge, we propose a goal-oriented clarification framework that aligns clarification behavior with ambiguity resolution. Central to our approach is the Information Gain Reward, a metric that quantifies the utility of clarification questions by measuring the Bayesian belief update towards the ground-truth goal induced by the clarification exchange. We train the clarifier (LLM) using this reward to optimize for high information gain, ensuring that clarifications effectively reduce uncertainty and improve task completion within the agent-tool-user environment. We validate our framework within a clarification-enhanced $τ$-Bench environment, conducting cross-agent evaluations across five heterogeneous backbones. Empirical results demonstrate that our method consistently improves the success rate by 3.7\% over the no-clarification baseline, while adding only 0.3 total interaction steps on average.

Rongwu Xu, Xuan Qi, Zehan Qi et al. · uw

The rise of large language models (LLMs) has enabled us to seek answers to inherently debatable questions on LLM chatbots, necessitating a reliable way to evaluate their ability. However, traditional QA benchmarks assume fixed answers are inadequate for this purpose. To address this, we introduce DebateQA, a dataset of 2,941 debatable questions, each accompanied by multiple human-annotated partial answers that capture a variety of perspectives. We develop two metrics: Perspective Diversity, which evaluates the comprehensiveness of perspectives, and Dispute Awareness, which assesses if the LLM acknowledges the question's debatable nature. Experiments demonstrate that both metrics align with human preferences and are stable across different underlying models. Using DebateQA with two metrics, we assess 12 popular LLMs and retrieval-augmented generation methods. Our findings reveal that while LLMs generally excel at recognizing debatable issues, their ability to provide comprehensive answers encompassing diverse perspectives varies considerably.

Xuming Hu, Junzhe Chen, Xiaochuan Li et al.

Large language models (LLMs) have recently driven striking performance improvements across a range of natural language processing tasks. The factual knowledge acquired during pretraining and instruction tuning can be useful in various downstream tasks, such as question answering, and language generation. Unlike conventional Knowledge Bases (KBs) that explicitly store factual knowledge, LLMs implicitly store facts in their parameters. Content generated by the LLMs can often exhibit inaccuracies or deviations from the truth, due to facts that can be incorrectly induced or become obsolete over time. To this end, we aim to comprehensively evaluate the extent and scope of factual knowledge within LLMs by designing the benchmark Pinocchio. Pinocchio contains 20K diverse factual questions that span different sources, timelines, domains, regions, and languages. Furthermore, we investigate whether LLMs are able to compose multiple facts, update factual knowledge temporally, reason over multiple pieces of facts, identify subtle factual differences, and resist adversarial examples. Extensive experiments on different sizes and types of LLMs show that existing LLMs still lack factual knowledge and suffer from various spurious correlations. We believe this is a critical bottleneck for realizing trustworthy artificial intelligence. The dataset Pinocchio and our codes will be publicly available.

Xuming Hu, Zhijiang Guo, Yu Fu et al.

A scene graph is a semantic representation that expresses the objects, attributes, and relationships between objects in a scene. Scene graphs play an important role in many cross modality tasks, as they are able to capture the interactions between images and texts. In this paper, we focus on scene graph modification (SGM), where the system is required to learn how to update an existing scene graph based on a natural language query. Unlike previous approaches that rebuilt the entire scene graph, we frame SGM as a graph expansion task by introducing the incremental structure expanding (ISE). ISE constructs the target graph by incrementally expanding the source graph without changing the unmodified structure. Based on ISE, we further propose a model that iterates between nodes prediction and edges prediction, inferring more accurate and harmonious expansion decisions progressively. In addition, we construct a challenging dataset that contains more complicated queries and larger scene graphs than existing datasets. Experiments on four benchmarks demonstrate the effectiveness of our approach, which surpasses the previous state-of-the-art model by large margins.

Tung Sum Thomas Kwok, Zeyong Zhang, Xinyu Wang et al.

Multi-step LLM reasoning over structured tables fails because planning and execution share no explicit cell-grounding contract. Existing methods constrain the planner to a left-to-right factorization at odds with table permutation invariance, and score intermediate states by generated content alone, overlooking cell grounding. We conduct a pilot study showing that diffusion language models (DLMs) produce more human-aligned and permutation-stable cell attention on tables than autoregressive models, with a 40.2% median reduction in attention-AUROC variability under row reordering. Motivated by this, we propose TABALIGN, a planned table reasoning framework that operationalizes the contract. TABALIGN pairs a masked DLM planner, whose bidirectional denoising emits plan steps as binary cell masks, with TABATTN, a lightweight verifier trained on 1,600 human-verified attention standards to score each step by its attention overlap with the plan-designated mask. Across eight benchmarks covering table question answering and fact verification, TABALIGN improves average accuracy by 15.76 percentage points over the strongest open-source baseline at comparable 8B-class scale, with a matched-backbone ablation attributing 2.87 percentage points of this gain to the DLM planner over an AR planner on a fixed reasoner. Cleaner DLM plans also accelerate downstream reasoning execution by 44.64%.

Ying Su, Mingwen Liu, Zhijiang Guo

Logical reasoning is a pivotal component in the field of artificial intelligence. Proof planning, particularly in contexts requiring the validation of explanation accuracy, continues to present challenges. The recent advancement of large language models (LLMs) has led to significant progress in natural language proof planning, evolving from one-stage generators to more complex three-stage systems that include additional searchers or verifiers. While these assisted methods improve the quality of generated results, they also introduce increased search efforts and computational costs. Furthermore, the generative process itself remains underexplored. In this study, we propose a stepwise decoding approach augmented by contrastive learning to address two common errors encountered during the LLM generator's decoding process. We fine-tune the language model using both vanilla and enhanced hard negatives to mitigate these decoding errors. Empirical results demonstrate the effectiveness of our strategy. Additionally, our further analysis reveals that even larger LLMs still struggle to generate rigorous logical chains.

Zhong-Zhi Li, Duzhen Zhang, Ming-Liang Zhang et al.

Achieving human-level intelligence requires refining the transition from the fast, intuitive System 1 to the slower, more deliberate System 2 reasoning. While System 1 excels in quick, heuristic decisions, System 2 relies on logical reasoning for more accurate judgments and reduced biases. Foundational Large Language Models (LLMs) excel at fast decision-making but lack the depth for complex reasoning, as they have not yet fully embraced the step-by-step analysis characteristic of true System 2 thinking. Recently, reasoning LLMs like OpenAI's o1/o3 and DeepSeek's R1 have demonstrated expert-level performance in fields such as mathematics and coding, closely mimicking the deliberate reasoning of System 2 and showcasing human-like cognitive abilities. This survey begins with a brief overview of the progress in foundational LLMs and the early development of System 2 technologies, exploring how their combination has paved the way for reasoning LLMs. Next, we discuss how to construct reasoning LLMs, analyzing their features, the core methods enabling advanced reasoning, and the evolution of various reasoning LLMs. Additionally, we provide an overview of reasoning benchmarks, offering an in-depth comparison of the performance of representative reasoning LLMs. Finally, we explore promising directions for advancing reasoning LLMs and maintain a real-time \href{https://github.com/zzli2022/Awesome-Slow-Reason-System}{GitHub Repository} to track the latest developments. We hope this survey will serve as a valuable resource to inspire innovation and drive progress in this rapidly evolving field.

Yebo Wu, Han Jin, Zhijiang Guo et al.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable reasoning capabilities yet continue to suffer from hallucination, where generated text contradicts visual content. In this paper, we introduce Dual-Anchor Introspective Decoding (DaID), a novel contrastive decoding framework that dynamically calibrates each token generation by mining the model's internal perceptual discrepancies. Specifically, DaID identifies a Spotlight layer to amplify visual factual signals and a Shadow layer to suppress textual inertia. By leveraging visual attention distributions to guide this dual-anchor selection process, our method ensures precise, token-specific adaptation. Experimental results across multiple benchmarks and MLLMs demonstrate that DaID significantly mitigates hallucination while enhancing general reasoning capabilities.

Hanxu Hu, Yuxuan Wang, Maggie Huan et al.

Reinforcement learning with Verifiable Rewards (RLVR) has emerged as a powerful paradigm for eliciting reasoning capabilities in large language models, particularly in mathematics and coding. While recent efforts have extended this paradigm to broader general scientific (STEM) domains, the complex interplay between supervised fine-tuning (SFT) and RL in these contexts remains underexplored. In this paper, we conduct controlled experiments revealing a critical challenge: for general STEM domains, RL applied directly to base models is highly sample-inefficient and is consistently surpassed by supervised fine-tuning (SFT) on moderate-quality responses. Yet sequential SFT followed by RL can further improve performance, suggesting that the two stages play complementary roles, and that how training data is allocated between them matters. Therefore, we propose DeReason, a difficulty-based data decoupling strategy for general reasoning. DeReason partitions training data by reasoning intensity estimated via LLM-based scoring into reasoning-intensive and non-reasoning-intensive subsets. It allocates broad-coverage, non-reasoning-intensive problems to SFT to establish foundational domain knowledge, and reserves a focused subset of difficult problems for RL to cultivate complex reasoning. We demonstrate that this principled decoupling yields better performance than randomly splitting the data for sequential SFT and RL. Extensive experiments on general STEM and mathematical benchmarks demonstrate that our decoupled curriculum training significantly outperforms SFT-only, RL-only, and random-split baselines. Our work provides a systematic study of the interplay between SFT and RL for general reasoning, offering a highly effective and generalized post-training recipe.

Zhicheng Yang, Zhijiang Guo, Yinya Huang et al.

Scaling test-time compute via long Chain-ofThought unlocks remarkable gains in reasoning capabilities, yet it faces practical limits due to the linear growth of KV cache and quadratic attention complexity. In this paper, we introduce Accordion-Thinking, an end-to-end framework where LLMs learn to self-regulate the granularity of the reasoning steps through dynamic summarization. This mechanism enables a Fold inference mode, where the model periodically summarizes its thought process and discards former thoughts to reduce dependency on historical tokens. We apply reinforcement learning to incentivize this capability further, uncovering a critical insight: the accuracy gap between the highly efficient Fold mode and the exhaustive Unfold mode progressively narrows and eventually vanishes over the course of training. This phenomenon demonstrates that the model learns to encode essential reasoning information into compact summaries, achieving effective compression of the reasoning context. Our Accordion-Thinker demonstrates that with learned self-compression, LLMs can tackle complex reasoning tasks with minimal dependency token overhead without compromising solution quality, and it achieves a 3x throughput while maintaining accuracy on a 48GB GPU memory configuration, while the structured step summaries provide a human-readable account of the reasoning process.

Yuxuan Yao, Haonan Sheng, Qingsong Lv et al.

The escalating scale of Large Language Models (LLMs) necessitates efficient adaptation techniques. Model merging has gained prominence for its efficiency and controllability. However, existing merging techniques typically serve as post-hoc refinements or focus on mitigating task interference, often failing to capture the dynamic optimization benefits of supervised fine-tuning (SFT). In this work, we propose Streaming Merging, an innovative model updating paradigm that conceptualizes merging as an iterative optimization process. Central to this paradigm is \textbf{ARM} (\textbf{A}ctivation-guided \textbf{R}otation-aware \textbf{M}erging), a strategy designed to approximate gradient descent dynamics. By treating merging coefficients as learning rates and deriving rotation vectors from activation subspaces, ARM effectively steers parameter updates along data-driven trajectories. Unlike conventional linear interpolation, ARM aligns semantic subspaces to preserve the geometric structure of high-dimensional parameter evolution. Remarkably, ARM requires only early SFT checkpoints and, through iterative merging, surpasses the fully converged SFT model. Experimental results across model scales (1.7B to 14B) and diverse domains (e.g., math, code) demonstrate that ARM can transcend converged checkpoints. Extensive experiments show that ARM provides a scalable and lightweight framework for efficient model adaptation.

Xinyu Zhou, Chang Jin, Carsten Eickhoff et al.

Large language models (LLMs) rarely admit uncertainty, often producing fluent but misleading answers, rather than abstaining (i.e., refusing to answer). This weakness is even evident in temporal question answering, where models frequently ignore time-sensitive evidence and conflate facts across different time-periods. In this paper, we present the first empirical study of training LLMs with an abstention ability while reasoning about temporal QA. Existing approaches such as calibration might be unreliable in capturing uncertainty in complex reasoning. We instead frame abstention as a teachable skill and introduce a pipeline that couples Chain-of-Thought (CoT) supervision with Reinforcement Learning (RL) guided by abstention-aware rewards. Our goal is to systematically analyze how different information types and training techniques affect temporal reasoning with abstention behavior in LLMs. Through extensive experiments studying various methods, we find that RL yields strong empirical gains on reasoning: a model initialized by Qwen2.5-1.5B-Instruct surpasses GPT-4o by $3.46\%$ and $5.80\%$ in Exact Match on TimeQA-Easy and Hard, respectively. Moreover, it improves the True Positive rate on unanswerable questions by $20\%$ over a pure supervised fine-tuned (SFT) variant. Beyond performance, our analysis shows that SFT induces overconfidence and harms reliability, while RL improves prediction accuracy but exhibits similar risks. Finally, by comparing implicit reasoning cues (e.g., original context, temporal sub-context, knowledge graphs) with explicit CoT supervision, we find that implicit information provides limited benefit for reasoning with abstention. Our study provides new insights into how abstention and reasoning can be jointly optimized, providing a foundation for building more reliable LLMs.

Haotian Xu, Xing Wu, Weinong Wang et al.

Can scaling transform reasoning? In this work, we explore the untapped potential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples, pioneering the development of a slow-thinking model, RedStar. Through extensive experiments with various LLMs and different sizes, we uncover the ingredients for specialization and scale for Long-CoT training. Surprisingly, even smaller models show significant performance gains with limited data, revealing the sample efficiency of Long-CoT and the critical role of sample difficulty in the learning process. Our findings demonstrate that Long-CoT reasoning can be effectively triggered with just a few thousand examples, while larger models achieve unparalleled improvements. We also introduce reinforcement learning (RL)-scale training as a promising direction for advancing slow-thinking systems. RedStar shines across domains: on the MATH-Hard benchmark, RedStar-code-math boosts performance from 66.2\% to 81.6\%, and on the USA Math Olympiad (AIME), it solves 46.7\% of problems using only 21k mixed-code-math datasets. In multimodal tasks like GeoQA and MathVista-GEO, RedStar-Geo achieves competitive results with minimal Long-CoT data, outperforming other slow-thinking systems like QvQ-Preview. Compared to QwQ, RedStar strikes the perfect balance between reasoning and generalizability. Our work highlights that, with careful tuning, scaling Long-CoT can unlock extraordinary reasoning capabilities-even with limited dataset and set a new standard for slow-thinking models across diverse challenges. Our data and models are released at https://huggingface.co/RedStar-Reasoning.

Jianbo Dai, Jianqiao Lu, Yunlong Feng et al.

Recent advancements in large language models (LLMs) have greatly improved code generation, specifically at the function level. For instance, GPT-4o has achieved a 91.0\% pass rate on HumanEval. However, this draws into question the adequacy of existing benchmarks in thoroughly assessing function-level code generation capabilities. Our study analyzed two common benchmarks, HumanEval and MBPP, and found that these might not thoroughly evaluate LLMs' code generation capacities due to limitations in quality, difficulty, and granularity. To resolve this, we introduce the Mostly Hard Python Problems (MHPP) dataset, consisting of 210 unique human-curated problems. By focusing on the combination of natural language and code reasoning, MHPP gauges LLMs' abilities to comprehend specifications and restrictions, engage in multi-step reasoning, and apply coding knowledge effectively. Initial evaluations of 26 LLMs using MHPP showed many high-performing models on HumanEval failed to achieve similar success on MHPP. Moreover, MHPP highlighted various previously undiscovered limitations within various LLMs, leading us to believe that it could pave the way for a better understanding of LLMs' capabilities and limitations. MHPP, evaluation pipeline, and leaderboard can be found in https://github.com/SparksofAGI/MHPP.

Yebo Wu, Jingguang Li, Chunlin Tian et al.

Federated fine-tuning enables privacy-preserving LLM adaptation but faces a critical bottleneck: the disparity between LLMs' high memory demands and edge devices' limited capacity. To break the memory barrier, we propose Chain Federated Fine-Tuning (ChainFed), an innovative paradigm that forgoes end-to-end updates in favor of a sequential, layer-by-layer manner. It first trains the initial adapter to convergence, freezes its weights, and then proceeds to the next. This iterative train-and-freeze process forms an optimization chain, gradually enhancing the model's task-specific proficiency. ChainFed further integrates three core techniques: 1) Dynamic Layer Co-Tuning to bridge semantic gaps between sequentially tuned layers and facilitate information flow; 2) Globally Perceptive Optimization to endow each adapter with foresight beyond its local objective; 3) Function-Oriented Adaptive Tuning to automatically identify the optimal fine-tuning starting point. Extensive experiments on multiple benchmarks demonstrate the superiority of ChainFed over existing methods, boosting average accuracy by up to 46.46\%.

Minrui Xu, Zilin Wang, Mengyi DENG et al.

Equipping LLMs with tool-use capabilities via Agentic Reinforcement Learning (Agentic RL) is bottlenecked by two challenges: the lack of scalable, robust execution environments and the scarcity of realistic training data that captures implicit human reasoning. Existing approaches depend on costly real-world APIs, hallucination-prone LLM simulators, or synthetic environments that are often single-turn or depend on pre-collected documents. Moreover, synthetic trajectories are frequently over-specified, resembling instruction sequences rather than natural human intents, reducing their effectiveness for RL training. We introduce EnvFactory, a fully automated framework that addresses both challenges. EnvFactory autonomously explores and verifies stateful, executable tool environments from authentic resources, and synthesizes natural multi-turn trajectories through topology-aware sampling and calibrated refinement, producing grounded queries with implicit intents. Using only 85 verified environments across 7 domains, EnvFactory generates 2,575 SFT and RL trajectories. Despite using significantly fewer environments than prior work, which are often 5 times more, EnvFactory achieves superior training efficiency and downstream performance, improving Qwen3-series models by up to +15% on BFCLv3, +8.6% on MCP-Atlas, and +6% on conversational benchmarks including $τ^2$-Bench and VitaBench. By fully automating both environment construction and trajectory synthesis, EnvFactory provides a scalable, extensible, and robust foundation for Agentic RL.

Zhicheng Yang, Zhijiang Guo, Yinya Huang et al.

Large Language Models (LLMs) have shown remarkable reasoning capabilities through Reinforcement Learning with Verifiable Rewards (RLVR) methods. However, a key limitation of existing approaches is that rewards defined at the full trajectory level provide insufficient guidance for optimizing the intermediate steps of a reasoning process. To address this, we introduce \textbf{\name}, a novel method that estimates the mathematical expectations of rewards at various reasoning steps using tree sampling. Unlike prior methods that rely on a separate step reward model, \name directly estimates these rewards through this sampling process. Building on the group-relative reward training mechanism of GRPO, \name innovatively computes rewards based on step-level groups generated during tree sampling. This advancement allows \name to produce fine-grained and dense reward signals, significantly enhancing the learning process and overall performance of LLMs. Experimental results demonstrate that our \name algorithm substantially improves the average Pass@1 accuracy of Qwen-2.5-Math on test benchmarks, increasing it from 19.0\% to 35.5\%. Furthermore, \name significantly outperforms GRPO by 2.9\% in performance while simultaneously reducing the average response length by 18.1\%, showcasing its effectiveness and efficiency. Our code will be available at \href{https://github.com/yangzhch6/TreeRPO}{https://github.com/yangzhch6/TreeRPO}.

Yuhao Qing, Boyu Zhu, Mingzhe Du et al. · mit

Existing code generation benchmarks primarily evaluate functional correctness, with limited focus on code efficiency and often restricted to a single language like Python. To address this gap, we introduce EffiBench-X, the first multi-language benchmark designed to measure the efficiency of LLM-generated code. EffiBench-X supports Python, C++, Java, JavaScript, Ruby, and Golang. It comprises competitive programming tasks with human-expert solutions as efficiency baselines. Evaluating state-of-the-art LLMs on EffiBench-X reveals that while models generate functionally correct code, they consistently underperform human experts in efficiency. Even the most efficient LLM-generated solutions (Qwen3-32B) achieve only around \textbf{62\%} of human efficiency on average, with significant language-specific variations. LLMs show better efficiency in Python, Ruby, and JavaScript than in Java, C++, and Golang. For instance, DeepSeek-R1's Python code is significantly more efficient than its Java code. These results highlight the critical need for research into LLM optimization techniques to improve code efficiency across diverse languages. The dataset and evaluation infrastructure are submitted and available at https://github.com/EffiBench/EffiBench-X.git and https://huggingface.co/datasets/EffiBench/effibench-x.

Jianqiao Lu, Yingjia Wan, Yinya Huang et al. · cambridge

Autoformalization aims to convert informal mathematical proofs into machine-verifiable formats, bridging the gap between natural and formal languages. However, ensuring semantic alignment between the informal and formalized statements remains challenging. Existing approaches heavily rely on manual verification, hindering scalability. To address this, we introduce \textsc{FormalAlign}, the first automated framework designed for evaluating the alignment between natural and formal languages in autoformalization. \textsc{FormalAlign} trains on both the autoformalization sequence generation task and the representational alignment between input and output, employing a dual loss that combines a pair of mutually enhancing autoformalization and alignment tasks. Evaluated across four benchmarks augmented by our proposed misalignment strategies, \textsc{FormalAlign} demonstrates superior performance. In our experiments, \textsc{FormalAlign} outperforms GPT-4, achieving an Alignment-Selection Score 11.58\% higher on \forml-Basic (99.21\% vs. 88.91\%) and 3.19\% higher on MiniF2F-Valid (66.39\% vs. 64.34\%). This effective alignment evaluation significantly reduces the need for manual verification. Both the dataset and code can be accessed via~\url{https://github.com/rookie-joe/FormalAlign}.

Yebo Wu, Chunlin Tian, Jingguang Li et al.

Large Language Models (LLMs) have demonstrated impressive success across various tasks. Integrating LLMs with Federated Learning (FL), a paradigm known as FedLLM, offers a promising avenue for collaborative model adaptation while preserving data privacy. This survey provides a systematic and comprehensive review of FedLLM. We begin by tracing the historical development of both LLMs and FL, summarizing relevant prior research to set the context. Subsequently, we delve into an in-depth analysis of the fundamental challenges inherent in deploying FedLLM. Addressing these challenges often requires efficient adaptation strategies; therefore, we conduct an extensive examination of existing Parameter-Efficient Fine-tuning (PEFT) methods and explore their applicability within the FL framework. To rigorously evaluate the performance of FedLLM, we undertake a thorough review of existing fine-tuning datasets and evaluation benchmarks. Furthermore, we discuss FedLLM's diverse real-world applications across multiple domains. Finally, we identify critical open challenges and outline promising research directions to foster future advancements in FedLLM. This survey aims to serve as a foundational resource for researchers and practitioners, offering valuable insights into the rapidly evolving landscape of federated fine-tuning for LLMs. It also establishes a roadmap for future innovations in privacy-preserving AI. We actively maintain a GitHub repo \href{https://github.com/Clin0212/Awesome-Federated-LLM-Learning}{https://github.com/Clin0212/Awesome-Federated-LLM-Learning} to track cutting-edge advancements in this field.

Yuan Chang, Ziyue Li, Hengyuan Zhang et al.

While Large Language Models (LLMs) have shown significant potential in assisting peer review, current methods often struggle to generate thorough and insightful reviews while maintaining efficiency. In this paper, we propose TreeReview, a novel framework that models paper review as a hierarchical and bidirectional question-answering process. TreeReview first constructs a tree of review questions by recursively decomposing high-level questions into fine-grained sub-questions and then resolves the question tree by iteratively aggregating answers from leaf to root to get the final review. Crucially, we incorporate a dynamic question expansion mechanism to enable deeper probing by generating follow-up questions when needed. We construct a benchmark derived from ICLR and NeurIPS venues to evaluate our method on full review generation and actionable feedback comments generation tasks. Experimental results of both LLM-based and human evaluation show that TreeReview outperforms strong baselines in providing comprehensive, in-depth, and expert-aligned review feedback, while reducing LLM token usage by up to 80% compared to computationally intensive approaches. Our code and benchmark dataset are available at https://github.com/YuanChang98/tree-review.

Dong Huang, Guangtao Zeng, Jianbo Dai et al.

As large language models (LLMs) play an increasingly important role in code generation, enhancing both correctness and efficiency has become crucial. Current methods primarily focus on correctness, often overlooking efficiency. To address this gap, we introduce EffiCoder to improve both aspects by fine-tuning LLMs on a high-quality dataset comprising correct and efficient code samples. Our methodology involves leveraging multiple LLMs to generate diverse candidate code solutions for various tasks across different programming languages. We then evaluate these solutions by measuring their execution time and memory usage through local execution. The code solution with the lowest execution time and memory consumption is selected as the final output for each task. Experimental results demonstrate significant improvements when fine-tuning with Effi-Instruct. For instance, Qwen2.5-Coder-7B-Instruct's pass@1 score increases from 44.8\% to 57.7\%, while the average execution time for correct tasks decreases by 48.4\%. EffiCoder offers a scalable and effective solution for advancing AI-driven code generation, benefiting software development and computational problem-solving. The source code of Effi-Code was released at https://github.com/huangd1999/EffiCoder.

Zhong-Zhi Li, Xiao Liang, Zihao Tang et al.

Large Language Models (LLMs) have recently achieved remarkable progress by leveraging Reinforcement Learning and extended Chain-of-Thought (CoT) techniques. However, the challenge of performing efficient language reasoning--especially during inference with extremely long outputs--has drawn increasing attention from the research community. In this work, we propose a dynamic ratio-based training pipeline that does not rely on sophisticated data annotations or interpolation between multiple models. We continuously balance the weights between the model's System-1 and System-2 data to eliminate redundant reasoning processes while preserving the model's reasoning capability. We validate our approach across models on DeepSeek-R1-Distill-7B and DeepSeek-R1-Distill-14B and on a diverse set of benchmarks with varying difficulty levels. Our method significantly reduces the number of output tokens by nearly 40% while maintaining the accuracy of the reasoning. Our code and data will be available soon.

Shaohang Wei, Wei Li, Feifan Song et al. · pku

Temporal reasoning is pivotal for Large Language Models (LLMs) to comprehend the real world. However, existing works neglect the real-world challenges for temporal reasoning: (1) intensive temporal information, (2) fast-changing event dynamics, and (3) complex temporal dependencies in social interactions. To bridge this gap, we propose a multi-level benchmark TIME, designed for temporal reasoning in real-world scenarios. TIME consists of 38,522 QA pairs, covering 3 levels with 11 fine-grained sub-tasks. This benchmark encompasses 3 sub-datasets reflecting different real-world challenges: TIME-Wiki, TIME-News, and TIME-Dial. We conduct extensive experiments on reasoning models and non-reasoning models. And we conducted an in-depth analysis of temporal reasoning performance across diverse real-world scenarios and tasks, and summarized the impact of test-time scaling on temporal reasoning capabilities. Additionally, we release TIME-Lite, a human-annotated subset to foster future research and standardized evaluation in temporal reasoning. The code is available at https://github.com/sylvain-wei/TIME , the dataset is available at https://huggingface.co/datasets/SylvainWei/TIME , and the project page link is https://sylvain-wei.github.io/TIME/ .

Fengwei Teng, Jinyi Bai, Xinhao Yao et al.

Group Relative Policy Optimization (GRPO) has proven effective in RLVR by using outcome-based rewards. While fine-grained dense rewards can theoretically improve performance, we reveal that under practical sampling budgets, Monte Carlo estimation yields high-variance and sign-inconsistent advantages for early reasoning tokens, paradoxically underperforming outcome-only GRPO. We propose Skip-Connected Optimization (SKPO), which decomposes reasoning into upstream and downstream phases: upstream receives dense rewards from downstream Monte Carlo sampling with single-stream optimization; downstream maintains group-relative optimization, where a skip connection concatenates the upstream segment with the original problem, enabling the model to leverage helpful upstream reasoning while preserving the freedom to bypass flawed reasoning through direct problem access. Experiments demonstrate improvements of 3.91% and 6.17% relative gains over the strongest baselines on Qwen2.5-Math-7B and Llama-3.2-3B respectively across mathematical benchmarks and out-of-domain tasks including general reasoning and code generation. Further analysis reveals an implicit advantage: SKPO generates trajectories with higher intermediate-step quality even when matched for final correctness.

Rongwu Xu, Zehan Qi, Zhijiang Guo et al. · uw

This survey provides an in-depth analysis of knowledge conflicts for large language models (LLMs), highlighting the complex challenges they encounter when blending contextual and parametric knowledge. Our focus is on three categories of knowledge conflicts: context-memory, inter-context, and intra-memory conflict. These conflicts can significantly impact the trustworthiness and performance of LLMs, especially in real-world applications where noise and misinformation are common. By categorizing these conflicts, exploring the causes, examining the behaviors of LLMs under such conflicts, and reviewing available solutions, this survey aims to shed light on strategies for improving the robustness of LLMs, thereby serving as a valuable resource for advancing research in this evolving area.

Mengyi Deng, Zhiwei Li, Xin Li et al.

Although Large Language Models (LLMs) have made remarkable progress, current preference optimization methods still struggle to align directional consistency while preserving reasoning diversity. To address this limitation, we propose Directional-Groupwise Preference Optimization (DGPO), a lightweight framework that aggregates supervision signals at the group level and explicitly models direction-aware alignment through multi-candidate comparisons. DGPO organizes forward and reverse question-answer instances into structured sets and optimizes a margin-based likelihood objective that separates coherent reasoning paths from inconsistent alternatives. This group-wise formulation captures richer relative information than pairwise objectives and reinforces consistency across diverse reasoning pathways. Empirical results show that our constructed reverse data yields a 3.2% average improvement across five benchmarks, while DGPO further delivers consistent gains across multiple datasets and model families, achieving average accuracy improvements of up to 3.6%.

Yinhong Liu, Han Zhou, Zhijiang Guo et al. · cambridge

Large Language Models (LLMs) have demonstrated promising capabilities as automatic evaluators in assessing the quality of generated natural language. However, LLMs still exhibit biases in evaluation and often struggle to generate coherent evaluations that align with human assessments. In this work, we first conduct a systematic study of the misalignment between LLM evaluators and human evaluation, revealing that existing calibration methods aimed at mitigating biases of LLMs are insufficient for effectively aligning LLM evaluators. Inspired by the use of preference data in RLHF, we formulate the evaluation as a ranking problem and introduce Pairwise-preference Search (PAIRS), an uncertainty-guided search-based rank aggregation method that employs LLMs to conduct pairwise comparisons locally and efficiently ranks candidate texts globally. PAIRS achieves state-of-the-art performance on representative evaluation tasks in long-form generations and demonstrates significant improvements over direct scoring. Furthermore, we provide insights into the role of pairwise preference in quantifying the transitivity of LLMs and demonstrate how PAIRS benefits from calibration using debiased pairwise evaluations.

Yingjia Wan, Haochen Tan, Xiao Zhu et al. · cambridge

Evaluating the factuality of long-form generations from Large Language Models (LLMs) remains challenging due to efficiency bottlenecks and reliability concerns. Prior efforts attempt this by decomposing text into claims, searching for evidence, and verifying claims, but suffer from critical drawbacks: (1) inefficiency due to overcomplicated pipeline components, and (2) ineffectiveness stemming from inaccurate claim sets and insufficient evidence. To address these limitations, we propose \textbf{FaStfact}, an evaluation framework that achieves the highest alignment with human evaluation and time/token efficiency among existing baselines. FaStfact first employs chunk-level claim extraction integrated with confidence-based pre-verification, significantly reducing the time and token cost while ensuring reliability. For searching and verification, it collects document-level evidence from crawled web-pages and selectively retrieves it during verification. Extensive experiments based on an annotated benchmark \textbf{FaStfact-Bench} demonstrate the reliability of FaStfact in both efficiently and effectively evaluating long-form factuality. Code, benchmark data, and annotation interface tool are available at https://github.com/Yingjia-Wan/FaStfact.

Jiale Guo, Suizhi Huang, Mei Li et al.

The integration of Large Language Models (LLMs) into software engineering has driven a transition from traditional rule-based systems to autonomous agentic systems capable of solving complex problems. However, systematic progress is hindered by a lack of comprehensive understanding of how benchmarks and solutions interconnect. This survey addresses this gap by providing the first holistic analysis of LLM-powered software engineering, offering insights into evaluation methodologies and solution paradigms. We review over 150 recent papers and propose a taxonomy along two key dimensions: (1) Solutions, categorized into prompt-based, fine-tuning-based, and agent-based paradigms, and (2) Benchmarks, including tasks such as code generation, translation, and repair. Our analysis highlights the evolution from simple prompt engineering to sophisticated agentic systems incorporating capabilities like planning, reasoning, memory mechanisms, and tool augmentation. To contextualize this progress, we present a unified pipeline illustrating the workflow from task specification to deliverables, detailing how different solution paradigms address various complexity levels. Unlike prior surveys that focus narrowly on specific aspects, this work connects 50+ benchmarks to their corresponding solution strategies, enabling researchers to identify optimal approaches for diverse evaluation criteria. We also identify critical research gaps and propose future directions, including multi-agent collaboration, self-evolving systems, and formal verification integration. This survey serves as a foundational guide for advancing LLM-driven software engineering. We maintain a GitHub repository that continuously updates the reviewed and related papers at https://github.com/lisaGuojl/LLM-Agent-SE-Survey.

Ruiran Su, Jiasheng Si, Zhijiang Guo et al.

Scientific fact-checking has mostly focused on text and tables, overlooking scientific charts, which are key for presenting quantitative evidence and statistical reasoning. We introduce ClimateViz, the first large-scale benchmark for scientific fact-checking using expert-curated scientific charts. ClimateViz contains 49,862 claims linked to 2,896 visualizations, each labeled as support, refute, or not enough information. To improve interpretability, each example includes structured knowledge graph explanations covering trends, comparisons, and causal relations. We evaluate state-of-the-art multimodal language models, including both proprietary and open-source systems, in zero-shot and few-shot settings. Results show that current models struggle with chart-based reasoning: even the best systems, such as Gemini 2.5 and InternVL 2.5, reach only 76.2 to 77.8 percent accuracy in label-only settings, far below human performance (89.3 and 92.7 percent). Explanation-augmented outputs improve performance in some models. We released our dataset and code alongside the paper.

Zhongshen Zeng, Yinhong Liu, Yingjia Wan et al.

Large language models (LLMs) have shown increasing capability in problem-solving and decision-making, largely based on the step-by-step chain-of-thought reasoning processes. However, evaluating these reasoning abilities has become increasingly challenging. Existing outcome-based benchmarks are beginning to saturate, becoming less effective in tracking meaningful progress. To address this, we present a process-based benchmark MR-Ben that demands a meta-reasoning skill, where LMs are asked to locate and analyse potential errors in automatically generated reasoning steps. Our meta-reasoning paradigm is especially suited for system-2 slow thinking, mirroring the human cognitive process of carefully examining assumptions, conditions, calculations, and logic to identify mistakes.MR-Ben comprises 5,975 questions curated by human experts across a wide range of subjects, including physics, chemistry, logic, coding, and more. Through our designed metrics for assessing meta-reasoning on this benchmark, we identify interesting limitations and weaknesses of current LLMs (open-source and closed-source models). For example, with models like the o1 series from OpenAI demonstrating strong performance by effectively scrutinizing the solution space, many other state-of-the-art models fall significantly behind on MR-Ben, exposing potential shortcomings in their training strategies and inference methodologies.

Jianqiao Lu, Yingjia Wan, Zhengying Liu et al.

Autoformalization, the conversion of natural language mathematics into formal languages, offers significant potential for advancing mathematical reasoning. However, existing efforts are limited to formal languages with substantial online corpora and struggle to keep pace with rapidly evolving languages like Lean 4. To bridge this gap, we propose a new benchmark \textbf{Form}alization for \textbf{L}ean~\textbf{4} (\textbf{\name}) designed to evaluate the autoformalization capabilities of large language models (LLMs). This benchmark encompasses a comprehensive assessment of questions, answers, formal statements, and proofs. Additionally, we introduce a \textbf{P}rocess-\textbf{S}upervised \textbf{V}erifier (\textbf{PSV}) model that leverages the precise feedback from Lean 4 compilers to enhance autoformalization. Our experiments demonstrate that the PSV method improves autoformalization, enabling higher accuracy using less filtered training data. Furthermore, when fine-tuned with data containing detailed process information, PSV can leverage the data more effectively, leading to more significant improvements in autoformalization for Lean 4. Our dataset and code are available at \url{https://github.com/rookie-joe/PDA}.

Jing Xiong, Zixuan Li, Chuanyang Zheng et al.

Recent advances in natural language processing, primarily propelled by Large Language Models (LLMs), have showcased their remarkable capabilities grounded in in-context learning. A promising avenue for guiding LLMs in intricate reasoning tasks involves the utilization of intermediate reasoning steps within the Chain-of-Thought (CoT) paradigm. Nevertheless, the central challenge lies in the effective selection of exemplars for facilitating in-context learning. In this study, we introduce a framework that leverages Dual Queries and Low-rank approximation Re-ranking (DQ-LoRe) to automatically select exemplars for in-context learning. Dual Queries first query LLM to obtain LLM-generated knowledge such as CoT, then query the retriever to obtain the final exemplars via both question and the knowledge. Moreover, for the second query, LoRe employs dimensionality reduction techniques to refine exemplar selection, ensuring close alignment with the input question's knowledge. Through extensive experiments, we demonstrate that DQ-LoRe significantly outperforms prior state-of-the-art methods in the automatic selection of exemplars for GPT-4, enhancing performance from 92.5% to 94.2%. Our comprehensive analysis further reveals that DQ-LoRe consistently outperforms retrieval-based approaches in terms of both performance and adaptability, especially in scenarios characterized by distribution shifts. DQ-LoRe pushes the boundary of in-context learning and opens up new avenues for addressing complex reasoning challenges. Our code is released at https://github.com/menik1126/DQ-LoRe

Chunlin Tian, Zhan Shi, Zhijiang Guo et al.

Adapting Large Language Models (LLMs) to new tasks through fine-tuning has been made more efficient by the introduction of Parameter-Efficient Fine-Tuning (PEFT) techniques, such as LoRA. However, these methods often underperform compared to full fine-tuning, particularly in scenarios involving complex datasets. This issue becomes even more pronounced in complex domains, highlighting the need for improved PEFT approaches that can achieve better performance. Through a series of experiments, we have uncovered two critical insights that shed light on the training and parameter inefficiency of LoRA. Building on these insights, we have developed HydraLoRA, a LoRA framework with an asymmetric structure that eliminates the need for domain expertise. Our experiments demonstrate that HydraLoRA outperforms other PEFT approaches, even those that rely on domain knowledge during the training and inference phases.

Zhicheng Yang, Zhijiang Guo, Yifan Song et al.

On-policy distillation (OPD) leverages dense teacher rewards to enhance reasoning models. However, scaling OPD to long-horizon tasks exposes a critical flaw: as the student's generated prefix inevitably diverges from the teacher's thought process, the teacher's dense reward loses local exploitability. Continuing to generate and evaluate tokens on these ``drifted'' trajectories not only degrades reward quality but also incurs massive computational waste. To address this, we introduce \textbf{Prune-OPD}, a framework that dynamically aligns training budgets with supervision quality. By continuously monitoring the local compatibility between student and teacher predictions (e.g., via top-$k$ overlap), Prune-OPD detects prefix-drift events in real time. Upon detecting severe drift, it monotonically down-weights subsequent unreliable rewards and triggers dynamic rollout truncation. This allows the training process to halt futile generation and reallocate compute strictly to reliable teacher supervision. Across diverse teacher-student combinations, Prune-OPD consistently aligns computation with supervision reliability. When prefix drift makes dense teacher rewards unreliable, it reduces training time by 37.6\%--68.0\% while preserving, and often improving, performance on challenging benchmarks (AMC, AIME, HMMT). When student-teacher compatibility remains high, it automatically preserves long-context supervision by expanding the training window. These results suggest that Prune-OPD improves OPD not by blindly shortening rollouts, but by reallocating computation toward locally exploitable teacher rewards.

Xinyu Zhou, Boyu Zhu, Haotian Zhang et al.

Reinforcement learning (RL) plays a central role in improving the reasoning and alignment of large language models, yet its efficiency critically depends on how training data are selected. Existing online selection strategies predominantly rely on difficulty-based heuristics, favouring datapoints with intermediate success rates, implicitly equating difficulty with informativeness and neglecting epistemic uncertainty arising from limited evidence. We introduce InSight, an INformation-guided data SamplInG metHod for RL Training, grounded in a weighted mutual information objective. By modeling data outcomes with Bayesian latent success rates, we show that expected uncertainty reduction decomposes into complementary difficulty- and evidence-dependent components, revealing a fundamental limitation of difficulty-only selection. Leveraging this observation, InSight constructs a stable acquisition score based on the mean belief of datapoints' success rather than noisy sampled outcomes, and naturally extends to multi-rollout settings common in reinforcement learning with verifiable rewards (RLVR). Extensive experiments demonstrate that InSight consistently achieves state-of-the-art performance and improves training efficiency, including a +1.41 average gain on Planning & Mathmatics benchmarks, +1.01 improvement on general reasoning, and up to ~2.2x acceleration, with negligible additional computational overhead.

Zehan Qi, Rongwu Xu, Zhijiang Guo et al. · uw

Retrieval-augmented generation (RAG) is a promising approach to address the limitations of fixed knowledge in large language models (LLMs). However, current benchmarks for evaluating RAG systems suffer from two key deficiencies: (1) they fail to adequately measure LLMs' capability in handling long-context retrieval due to a lack of datasets that reflect the characteristics of retrieved documents, and (2) they lack a comprehensive evaluation method for assessing LLMs' ability to generate long-form responses that effectively exploits retrieved information. To address these shortcomings, we introduce the Long$^2$RAG benchmark and the Key Point Recall (KPR) metric. Long$^2$RAG comprises 280 questions spanning 10 domains and across 8 question categories, each associated with 5 retrieved documents with an average length of 2,444 words. KPR evaluates the extent to which LLMs incorporate key points extracted from the retrieved documents into their generated responses, providing a more nuanced assessment of their ability to exploit retrieved information.

Xiao Liang, Zhongzhi Li, Yeyun Gong et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has recently emerged as a key paradigm for post-training Large Language Models (LLMs), particularly for complex reasoning tasks. However, vanilla RLVR training has been shown to improve Pass@1 performance at the expense of policy entropy, leading to reduced generation diversity and limiting the Pass@k performance, which typically represents the upper bound of LLM reasoning capability. In this paper, we systematically analyze the policy's generation diversity from the perspective of training problems and find that augmenting and updating training problems helps mitigate entropy collapse during training. Based on these observations, we propose an online Self-play with Variational problem Synthesis (SvS) strategy for RLVR training, which uses the policy's correct solutions to synthesize variational problems while ensuring their reference answers remain identical to the originals. This self-improving strategy effectively maintains policy entropy during training and substantially improves Pass@k compared with standard RLVR, sustaining prolonged improvements and achieving absolute gains of 18.3% and 22.8% in Pass@32 performance on the competition-level AIME24 and AIME25 benchmarks. Experiments on 12 reasoning benchmarks across varying model sizes from 3B to 32B consistently demonstrate the generalizability and robustness of SvS.

Michael Schlichtkrull, Yulong Chen, Chenxi Whitehouse et al. · amazon-science

The Automated Verification of Textual Claims (AVeriTeC) shared task asks participants to retrieve evidence and predict veracity for real-world claims checked by fact-checkers. Evidence can be found either via a search engine, or via a knowledge store provided by the organisers. Submissions are evaluated using AVeriTeC score, which considers a claim to be accurately verified if and only if both the verdict is correct and retrieved evidence is considered to meet a certain quality threshold. The shared task received 21 submissions, 18 of which surpassed our baseline. The winning team was TUDA_MAI with an AVeriTeC score of 63%. In this paper we describe the shared task, present the full results, and highlight key takeaways from the shared task.

Caiqi Zhang, Zhijiang Guo, Andreas Vlachos · cambridge

This paper investigates the potential benefits of language-specific fact-checking models, focusing on the case of Chinese. We first demonstrate the limitations of translation-based methods and multilingual large language models (e.g., GPT-4), highlighting the need for language-specific systems. We further propose a Chinese fact-checking system that can better retrieve evidence from a document by incorporating context information. To better analyze token-level biases in different systems, we construct an adversarial dataset based on the CHEF dataset, where each instance has large word overlap with the original one but holds the opposite veracity label. Experimental results on the CHEF dataset and our adversarial dataset show that our proposed method outperforms translation-based methods and multilingual LLMs and is more robust toward biases, while there is still large room for improvement, emphasizing the importance of language-specific fact-checking systems.

Yuxuan Yao, Han Wu, Zhijiang Guo et al.

Large language models (LLMs) have demonstrated outstanding performance across various tasks, yet they still exhibit limitations such as hallucination, unfaithful reasoning, and toxic content. One potential approach to mitigate these issues is learning from human or external feedback (e.g. tools). In this paper, we introduce an intrinsic self-correct reasoning framework for LLMs that eliminates the need for human feedback, external tools, and handcraft prompts. The proposed framework, based on a multi-step reasoning paradigm \textbf{Le}arning from \textbf{Co}rrectness (\textsc{LeCo}), improves reasoning performance without needing to learn from errors. This paradigm prioritizes learning from correct reasoning steps, and a unique method to measure confidence for each reasoning step based on generation logits. Experimental results across various multi-step reasoning tasks demonstrate the effectiveness of the framework in improving reasoning performance with reduced token consumption.

Jianqiao Lu, Wanjun Zhong, Yufei Wang et al.

Although large language models (LLMs) have demonstrated adeptness in a range of tasks, they still lag behind human learning efficiency. This disparity is often linked to the inherent human capacity to learn from basic examples, gradually generalize and handle more complex problems, and refine their skills with continuous feedback. Inspired by this, this paper introduces YODA, a novel teacher-student progressive learning framework that emulates the teacher-student education process to improve the efficacy of model fine-tuning. The framework operates on an interactive \textit{basic-generalized-harder} loop. The teacher agent provides tailored feedback on the student's answers, and systematically organizes the education process. This process unfolds by teaching the student basic examples, reinforcing understanding through generalized questions, and then enhancing learning by posing questions with progressively enhanced complexity. With the teacher's guidance, the student learns to iteratively refine its answer with feedback, and forms a robust and comprehensive understanding of the posed questions. The systematic procedural data, which reflects the progressive learning process of humans, is then utilized for model training. Taking math reasoning as a testbed, experiments show that training LLaMA2 with data from YODA improves SFT with significant performance gain (+17.01\% on GSM8K and +9.98\% on MATH). In addition, we find that training with curriculum learning further improves learning robustness.