Jiaxi Yang

Jiaxi Yang, Binyuan Hui, Min Yang et al.

The capability gap between open-source and closed-source large language models (LLMs) remains a challenge in text-to-SQL tasks. In this paper, we introduce a synthetic data approach that combines data produced by larger, more powerful models (strong models) with error information data generated by smaller, not well-aligned models (weak models). The method not only enhances the domain generalization of text-to-SQL models but also explores the potential of error data supervision through preference learning. Furthermore, we employ the synthetic data approach for instruction tuning on open-source LLMs, resulting SENSE, a specialized text-to-SQL model. The effectiveness of SENSE is demonstrated through state-of-the-art results on the SPIDER and BIRD benchmarks, bridging the performance gap between open-source models and methods prompted by closed-source models.

Binyuan Hui, Jian Yang, Zeyu Cui et al.

In this report, we introduce the Qwen2.5-Coder series, a significant upgrade from its predecessor, CodeQwen1.5. This series includes six models: Qwen2.5-Coder-(0.5B/1.5B/3B/7B/14B/32B). As a code-specific model, Qwen2.5-Coder is built upon the Qwen2.5 architecture and continues pretrained on a vast corpus of over 5.5 trillion tokens. Through meticulous data cleaning, scalable synthetic data generation, and balanced data mixing, Qwen2.5-Coder demonstrates impressive code generation capabilities while retaining general and math skills. These models have been evaluated on a wide range of code-related tasks, achieving state-of-the-art (SOTA) performance across more than 10 benchmarks, including code generation, completion, reasoning, and repair, consistently outperforming larger models of the same model size. We believe that the release of the Qwen2.5-Coder series will advance research in code intelligence and, with its permissive licensing, support wider adoption by developers in real-world applications.

Yang Zhao, Jiaxi Yang, Wenbo Wang et al.

Industrial systems demand reliable predictive maintenance strategies to enhance operational efficiency and reduce downtime. This paper introduces an integrated framework that leverages the capabilities of the Transformer model-based neural networks and deep reinforcement learning (DRL) algorithms to optimize system maintenance actions. Our approach employs the Transformer model to effectively capture complex temporal patterns in sensor data, thereby accurately predicting the remaining useful life (RUL) of an equipment. Additionally, the DRL component of our framework provides cost-effective and timely maintenance recommendations. We validate the efficacy of our framework on the NASA C-MPASS dataset, where it demonstrates significant advancements in both RUL prediction accuracy and the optimization of maintenance actions, compared to the other prevalent machine learning-based methods. Our proposed approach provides an innovative data-driven framework for industry machine systems, accurately forecasting equipment lifespans and optimizing maintenance schedules, thereby reducing downtime and cutting costs.

KaShun Shum, Binyuan Hui, Jiawei Chen et al.

Execution-based feedback like unit testing is widely used in the development of coding agents through test-time scaling (TTS) and reinforcement learning (RL). This paradigm requires scalable and reliable collection of unit test cases to provide accurate feedback, and the resulting feedback is often sparse and cannot effectively distinguish between trajectories that are both successful or both unsuccessful. In contrast, execution-free feedback from reward models can provide more fine-grained signals without depending on unit test cases. Despite this potential, execution-free feedback for realistic software engineering (SWE) agents remains underexplored. Aiming to develop versatile reward models that are effective across TTS and RL, however, we observe that two verifiers with nearly identical TTS performance can nevertheless yield very different results in RL. Intuitively, TTS primarily reflects the model's ability to select the best trajectory, but this ability does not necessarily generalize to RL. To address this limitation, we identify two additional aspects that are crucial for RL training: classification accuracy and calibration. We then conduct comprehensive controlled experiments to investigate how to train a robust reward model that performs well across these metrics. In particular, we analyze the impact of various factors such as training data scale, policy mixtures, and data source composition. Guided by these investigations, we introduce SWE-RM, an accurate and robust reward model adopting a mixture-of-experts architecture with 30B total parameters and 3B activated during inference. SWE-RM substantially improves SWE agents on both TTS and RL performance. For example, it increases the accuracy of Qwen3-Coder-Flash from 51.6% to 62.0%, and Qwen3-Coder-Max from 67.0% to 74.6% on SWE-Bench Verified using TTS, achieving new state-of-the-art performance among open-source models.

Jiajun Zhang, Zeyu Cui, Lei Zhang et al.

Code completion has become a central task, gaining significant attention with the rise of large language model (LLM)-based tools in software engineering. Although recent advances have greatly improved LLMs' code completion abilities, evaluation methods have not advanced equally. Most current benchmarks focus solely on functional correctness of code completions based on given context, overlooking models' ability to follow user instructions during completion-a common scenario in LLM-assisted programming. To address this limitation, we present the first instruction-guided code completion benchmark, Controllable Code Completion Benchmark (C3-Bench), comprising 2,195 carefully designed completion tasks. Through comprehensive evaluation of over 40 mainstream LLMs across C3-Bench and conventional benchmarks, we reveal substantial gaps in instruction-following capabilities between open-source and advanced proprietary models during code completion tasks. Moreover, we develop a straightforward data synthesis pipeline that leverages Qwen2.5-Coder to generate high-quality instruction-completion pairs for supervised fine-tuning (SFT). The resulting model, Qwen2.5-Coder-C3, achieves state-of-the-art performance on C3-Bench. Our findings provide valuable insights for enhancing LLMs' code completion and instruction-following capabilities, establishing new directions for future research in code LLMs. To facilitate reproducibility and foster further research in code LLMs, we open-source all code, datasets, and models.

Congzhen Shi, Ryan Rezai, Jiaxi Yang et al.

The rapid advancement of foundation models in medical imaging represents a significant leap toward enhancing diagnostic accuracy and personalized treatment. However, the deployment of foundation models in healthcare necessitates a rigorous examination of their trustworthiness, encompassing privacy, robustness, reliability, explainability, and fairness. The current body of survey literature on foundation models in medical imaging reveals considerable gaps, particularly in the area of trustworthiness. Additionally, existing surveys on the trustworthiness of foundation models do not adequately address their specific variations and applications within the medical imaging domain. This survey aims to fill that gap by presenting a novel taxonomy of foundation models used in medical imaging and analyzing the key motivations for ensuring their trustworthiness. We review current research on foundation models in major medical imaging applications, focusing on segmentation, medical report generation, medical question and answering (Q\&A), and disease diagnosis. These areas are highlighted because they have seen a relatively mature and substantial number of foundation models compared to other applications. We focus on literature that discusses trustworthiness in medical image analysis manuscripts. We explore the complex challenges of building trustworthy foundation models for each application, summarizing current concerns and strategies for enhancing trustworthiness. Furthermore, we examine the potential of these models to revolutionize patient care. Our analysis underscores the imperative for advancing towards trustworthy AI in medical image analysis, advocating for a balanced approach that fosters innovation while ensuring ethical and equitable healthcare delivery.

Mouxiang Chen, Lei Zhang, Yunlong Feng et al.

We propose SWE-Universe, a scalable and efficient framework for automatically constructing real-world software engineering (SWE) verifiable environments from GitHub pull requests (PRs). To overcome the prevalent challenges of automatic building, such as low production yield, weak verifiers, and prohibitive cost, our framework utilizes a building agent powered by an efficient custom-trained model. This agent employs iterative self-verification and in-loop hacking detection to ensure the reliable generation of high-fidelity, verifiable tasks. Using this method, we scale the number of real-world multilingual SWE environments to a million scale (807,693). We demonstrate the profound value of our environments through large-scale agentic mid-training and reinforcement learning. Finally, we applied this technique to Qwen3-Max-Thinking and achieved a score of 75.3% on SWE-Bench Verified. Our work provides both a critical resource and a robust methodology to advance the next generation of coding agents.

Yuheng Jing, Kai Li, Ziwen Zhang et al.

In-Context Reinforcement Learning (ICRL) has enabled foundation agents to adapt instantaneously to novel tasks, yet its efficacy in Ad-Hoc Teamwork (AHT)-where coordination with unknown partners is required-remains unexplored. To rigorously evaluate this, we introduce a large-scale benchmark ICRL4AHT, built upon a high-throughput JAX implementation of Overcooked-V2. Our benchmark includes a large, diverse teammate suite spanning both RL and heuristic policies, enabling controlled train-test shifts, and provides a reproducible end-to-end pipeline for teammate generation, learning-history collection, dataset construction, and online multi-episode evaluation. We evaluate representative history-conditioned ICRL algorithms, including Algorithm Distillation (AD) and Decision-Pretrained Transformer (DPT), across millions of transitions. Results reveal notable limitations: contrary to their success in single-agent domains, these baselines fail to exhibit robust test-time adaptation in multi-agent settings. Specifically, these methods frequently underperform random baselines across both unseen teammate and unseen layout tracks, with no clear in-context improvement over long horizons. These findings highlight the challenges of strategic inference under partial observability within the OvercookedV2 AHT protocol, establishing our benchmark as a critical testbed for next-generation coordination algorithms.

Shanghaoran Quan, Jiaxi Yang, Bowen Yu et al.

With the increasing code reasoning capabilities of existing large language models (LLMs) and breakthroughs in reasoning models like OpenAI o1 and o3, there is a growing need to develop more challenging and comprehensive benchmarks that effectively test their sophisticated competition-level coding abilities. Existing benchmarks, like LiveCodeBench and USACO, fall short due to the unavailability of private test cases, lack of support for special judges, and misaligned execution environments. To bridge this gap, we introduce CodeElo, a standardized competition-level code generation benchmark that effectively addresses all these challenges for the first time. CodeElo benchmark is mainly based on the official CodeForces platform and tries to align with the platform as much as possible. We compile the recent six months of contest problems on CodeForces with detailed information such as contest divisions, problem difficulty ratings, and problem algorithm tags. We introduce a unique judging method in which problems are submitted directly to the platform and develop a reliable Elo rating calculation system that aligns with the platform and is comparable with human participants but has lower variance. By testing on our CodeElo, we provide the Elo ratings of 30 existing popular open-source and 3 proprietary LLMs for the first time. The results show that o1-mini and QwQ-32B-Preview stand out significantly, achieving Elo ratings of 1578 and 1261, respectively, while other models struggle even with the easiest problems, placing in the lowest 25 percent among all human participants. Detailed analysis experiments are also conducted to provide insights into performance across algorithms and comparisons between using C++ and Python, which can suggest directions for future studies.

Jiaxi Yang, Wenglong Deng, Benlin Liu et al.

Data valuation plays a crucial role in machine learning. Existing data valuation methods, mainly focused on discriminative models, overlook generative models that have gained attention recently. In generative models, data valuation measures the impact of training data on generated datasets. Very few existing attempts at data valuation methods designed for deep generative models either concentrate on specific models or lack robustness in their outcomes. Moreover, efficiency still reveals vulnerable shortcomings. We formulate the data valuation problem in generative models from a similarity matching perspective to bridge the gaps. Specifically, we introduce Generative Model Valuator (GMValuator), the first training-free and model-agnostic approach to providing data valuation for image generation tasks. It empowers efficient data valuation through our innovative similarity matching module, calibrates biased contributions by incorporating image quality assessment, and attributes credits to all training samples based on their contributions to the generated samples. Additionally, we introduce four evaluation criteria for assessing data valuation methods in generative models. GMValuator is extensively evaluated on benchmark and high-resolution datasets and various mainstream generative architectures to demonstrate its effectiveness.

Jian Yang, Jiaxi Yang, Ke Jin et al.

Code large language models (codeLLMs) have made significant strides in code generation. Most previous code-related benchmarks, which consist of various programming exercises along with the corresponding test cases, are used as a common measure to evaluate the performance and capabilities of code LLMs. However, the current code LLMs focus on synthesizing the correct code snippet, ignoring the alignment with human preferences, where the query should be sampled from the practical application scenarios and the model-generated responses should satisfy the human preference. To bridge the gap between the model-generated response and human preference, we present a rigorous human-curated benchmark CodeArena to emulate the complexity and diversity of real-world coding tasks, where 397 high-quality samples spanning 40 categories and 44 programming languages, carefully curated from user queries. Further, we propose a diverse synthetic instruction corpus SynCode-Instruct (nearly 20B tokens) by scaling instructions from the website to verify the effectiveness of the large-scale synthetic instruction fine-tuning, where Qwen2.5-SynCoder totally trained on synthetic instruction data can achieve top-tier performance of open-source code LLMs. The results find performance differences between execution-based benchmarks and CodeArena. Our systematic experiments of CodeArena on 40+ LLMs reveal a notable performance gap between open SOTA code LLMs (e.g. Qwen2.5-Coder) and proprietary LLMs (e.g., OpenAI o1), underscoring the importance of the human preference alignment.\footnote{\url{https://codearenaeval.github.io/ }}

Jian Yang, Jiajun Zhang, Jiaxi Yang et al.

Code completion has become an essential tool for daily software development. Existing evaluation benchmarks often employ static methods that do not fully capture the dynamic nature of real-world coding environments and face significant challenges, including limited context length, reliance on superficial evaluation metrics, and potential overfitting to training datasets. In this work, we introduce a novel framework for enhancing code completion in software development through the creation of a repository-level benchmark ExecRepoBench and the instruction corpora Repo-Instruct, aim at improving the functionality of open-source large language models (LLMs) in real-world coding scenarios that involve complex interdependencies across multiple files. ExecRepoBench includes 1.2K samples from active Python repositories. Plus, we present a multi-level grammar-based completion methodology conditioned on the abstract syntax tree to mask code fragments at various logical units (e.g. statements, expressions, and functions). Then, we fine-tune the open-source LLM with 7B parameters on Repo-Instruct to produce a strong code completion baseline model Qwen2.5-Coder-Instruct-C based on the open-source model. Qwen2.5-Coder-Instruct-C is rigorously evaluated against existing benchmarks, including MultiPL-E and ExecRepoBench, which consistently outperforms prior baselines across all programming languages. The deployment of \ourmethod{} can be used as a high-performance, local service for programming development\footnote{\url{https://execrepobench.github.io/}}.

Qiyue Gao, Xinyu Pi, Kevin Liu et al. · cmu

Internal world models (WMs) enable agents to understand the world's state and predict transitions, serving as the basis for advanced deliberative reasoning. Recent large Vision-Language Models (VLMs), such as OpenAI o3, GPT-4o and Gemini, exhibit potential as general-purpose WMs. While the latest studies have evaluated and shown limitations in specific capabilities such as visual understanding, a systematic evaluation of VLMs' fundamental WM abilities remains absent. Drawing on comparative psychology and cognitive science, we propose a two-stage framework that assesses Perception (visual, spatial, temporal, quantitative, and motion) and Prediction (mechanistic simulation, transitive inference, compositional inference) to provide an atomic evaluation of VLMs as WMs. Guided by this framework, we introduce WM-ABench, a large-scale benchmark comprising 23 fine-grained evaluation dimensions across 6 diverse simulated environments with controlled counterfactual simulations. Through 660 experiments on 15 latest commercial and open-source VLMs, we find that these models exhibit striking limitations in basic world modeling abilities. For instance, almost all models perform at near-random accuracy when distinguishing motion trajectories. Additionally, they lack disentangled understanding -- e.g., some models tend to believe blue objects move faster than green ones. More rich results and analyses reveal significant gaps between VLMs and human-level world modeling.

An Yang, Anfeng Li, Baosong Yang et al. · tsinghua

In this work, we present Qwen3, the latest version of the Qwen model family. Qwen3 comprises a series of large language models (LLMs) designed to advance performance, efficiency, and multilingual capabilities. The Qwen3 series includes models of both dense and Mixture-of-Expert (MoE) architectures, with parameter scales ranging from 0.6 to 235 billion. A key innovation in Qwen3 is the integration of thinking mode (for complex, multi-step reasoning) and non-thinking mode (for rapid, context-driven responses) into a unified framework. This eliminates the need to switch between different models--such as chat-optimized models (e.g., GPT-4o) and dedicated reasoning models (e.g., QwQ-32B)--and enables dynamic mode switching based on user queries or chat templates. Meanwhile, Qwen3 introduces a thinking budget mechanism, allowing users to allocate computational resources adaptively during inference, thereby balancing latency and performance based on task complexity. Moreover, by leveraging the knowledge from the flagship models, we significantly reduce the computational resources required to build smaller-scale models, while ensuring their highly competitive performance. Empirical evaluations demonstrate that Qwen3 achieves state-of-the-art results across diverse benchmarks, including tasks in code generation, mathematical reasoning, agent tasks, etc., competitive against larger MoE models and proprietary models. Compared to its predecessor Qwen2.5, Qwen3 expands multilingual support from 29 to 119 languages and dialects, enhancing global accessibility through improved cross-lingual understanding and generation capabilities. To facilitate reproducibility and community-driven research and development, all Qwen3 models are publicly accessible under Apache 2.0.

Qwen, An Yang, Baosong Yang et al.

In this report, we introduce Qwen2.5, a comprehensive series of large language models (LLMs) designed to meet diverse needs. Compared to previous iterations, Qwen 2.5 has been significantly improved during both the pre-training and post-training stages. In terms of pre-training, we have scaled the high-quality pre-training datasets from the previous 7 trillion tokens to 18 trillion tokens. This provides a strong foundation for common sense, expert knowledge, and reasoning capabilities. In terms of post-training, we implement intricate supervised finetuning with over 1 million samples, as well as multistage reinforcement learning. Post-training techniques enhance human preference, and notably improve long text generation, structural data analysis, and instruction following. To handle diverse and varied use cases effectively, we present Qwen2.5 LLM series in rich sizes. Open-weight offerings include base and instruction-tuned models, with quantized versions available. In addition, for hosted solutions, the proprietary models currently include two mixture-of-experts (MoE) variants: Qwen2.5-Turbo and Qwen2.5-Plus, both available from Alibaba Cloud Model Studio. Qwen2.5 has demonstrated top-tier performance on a wide range of benchmarks evaluating language understanding, reasoning, mathematics, coding, human preference alignment, etc. Specifically, the open-weight flagship Qwen2.5-72B-Instruct outperforms a number of open and proprietary models and demonstrates competitive performance to the state-of-the-art open-weight model, Llama-3-405B-Instruct, which is around 5 times larger. Qwen2.5-Turbo and Qwen2.5-Plus offer superior cost-effectiveness while performing competitively against GPT-4o-mini and GPT-4o respectively. Additionally, as the foundation, Qwen2.5 models have been instrumental in training specialized models such as Qwen2.5-Math, Qwen2.5-Coder, QwQ, and multimodal models.

Lei Zhang, Yunshui Li, Ziqiang Liu et al.

With the advancement of large language models (LLMs) and the expansion of their context windows, existing long-context benchmarks fall short in effectively evaluating the models' comprehension and reasoning abilities in extended texts. Moreover, conventional benchmarks relying on F1 metrics often inaccurately score responses: they may undervalue correct answers that differ from the reference responses and overvalue incorrect ones that resemble the reference texts. In response to these limitations, we introduce Marathon, a novel evaluation benchmark that adopts a multiple-choice question format. It is specifically designed to overcome the constraints of previous benchmarks and provide a rapid, precise, and unbiased appraisal of the long-context comprehension skills of large language models. We conducted comprehensive evaluations on the Marathon benchmark with a range of state-of-the-art LLMs and assessed the effectiveness of various optimization strategies tailored for long-context generation. We anticipate that the Marathon benchmark and its associated leaderboard will enable a more precise and equitable evaluation of LLMs' capabilities in understanding and reasoning over extended contexts. Marathon is available at https://github.com/Hambaobao/Marathon.

Zeyao Ma, Jing Zhang, Xiaokang Zhang et al.

Large language models (LLMs) have demonstrated strong coding capabilities but still struggle to solve competitive programming problems correctly in a single attempt. Execution-based re-ranking offers a promising test-time scaling strategy, yet existing methods are constrained by either difficult test case generation or inefficient random input sampling. To address this limitation, we propose Agentic Verifier, an execution-based agent that actively reasons about program behaviors and searches for highly discriminative test inputs that expose behavioral discrepancies among candidate solutions. Through multi-turn interaction with code execution environments, the verifier iteratively refines the candidate input generator and produces targeted counterexamples rather than blindly sampling inputs. We train the verifier to acquire this discriminative input generation capability via a scalable pipeline combining large-scale data synthesis, rejection fine-tuning, and agentic reinforcement learning. Extensive experiments across five competitive programming benchmarks demonstrate consistent improvements over strong execution-based baselines, achieving up to +10-15% absolute gains in Best@K accuracy. Further analysis reveals clear test-time scaling behavior and highlights the verifier's broader potential beyond reranking.

Jiaxi Yang, Zihao Guo, Sheng Cao et al.

Auction-based Federated Learning (AFL) enables open collaboration among self-interested data consumers and data owners. Existing AFL approaches are commonly under the assumption of sellers' market in that the service clients as sellers are treated as scarce resources so that the aggregation servers as buyers need to compete the bids. Yet, as the technology progresses, an increasing number of qualified clients are now capable of performing federated learning tasks, leading to shift from sellers' market to a buyers' market. In this paper, we shift the angle by adapting the procurement auction framework, aiming to explain the pricing behavior under buyers' market. Our modeling starts with basic setting under complete information, then move further to the scenario where sellers' information are not fully observable. In order to select clients with high reliability and data quality, and to prevent from external attacks, we utilize a blockchain-based reputation mechanism. The experimental results validate the effectiveness of our approach.

Yang Zhao, Jiaxi Yang, Yiling Tao et al.

The increasing demand for privacy-preserving machine learning has spurred interest in federated unlearning, which enables the selective removal of data from models trained in federated systems. However, developing federated unlearning methods presents challenges, particularly in balancing three often conflicting objectives: privacy, accuracy, and efficiency. This paper provides a comprehensive analysis of existing federated unlearning approaches, examining their algorithmic efficiency, impact on model accuracy, and effectiveness in preserving privacy. We discuss key trade-offs among these dimensions and highlight their implications for practical applications across various domains. Additionally, we propose the OpenFederatedUnlearning framework, a unified benchmark for evaluating federated unlearning methods, incorporating classic baselines and diverse performance metrics. Our findings aim to guide practitioners in navigating the complex interplay of these objectives, offering insights to achieve effective and efficient federated unlearning. Finally, we outline directions for future research to further advance the state of federated unlearning techniques.

Liang Zhu, Yuelin Bai, Xiankun Ren et al.

Preference learning in Large Language Models (LLMs) has advanced significantly, yet existing methods remain limited by modest performance gains, high computational costs, hyperparameter sensitivity, and insufficient modeling of global token-level relationships. We introduce PLOT, which enhances Preference Learning in fine-tuning-based alignment through a token-level loss derived from Optimal Transport. By formulating preference learning as an Optimal Transport Problem, PLOT aligns model outputs with human preferences while preserving the original distribution of LLMs, ensuring stability and robustness. Furthermore, PLOT leverages token embeddings to capture semantic relationships, enabling globally informed optimization. Experiments across two preference categories - Human Values and Logic & Problem Solving - spanning seven subpreferences demonstrate that PLOT consistently improves alignment performance while maintaining fluency and coherence. These results substantiate optimal transport as a principled methodology for preference learning, establishing a theoretically grounded framework that provides new insights for preference learning of LLMs.

Lei Zhang, Jiaxi Yang, Min Yang et al.

We introduce **SWE-Flow**, a novel data synthesis framework grounded in Test-Driven Development (TDD). Unlike existing software engineering data that rely on human-submitted issues, **SWE-Flow** automatically infers incremental development steps directly from unit tests, which inherently encapsulate high-level requirements. The core of **SWE-Flow** is the construction of a Runtime Dependency Graph (RDG), which precisely captures function interactions, enabling the generation of a structured, step-by-step *development schedule*. At each step, **SWE-Flow** produces a partial codebase, the corresponding unit tests, and the necessary code modifications, resulting in fully verifiable TDD tasks. With this approach, we generated 16,061 training instances and 2,020 test instances from real-world GitHub projects, creating the **SWE-Flow-Eval** benchmark. Our experiments show that fine-tuning open model on this dataset significantly improves performance in TDD-based coding. To facilitate further research, we release all code, datasets, models, and Docker images at [Github](https://github.com/Hambaobao/SWE-Flow).

Ziwen Chen, Zhendong Wang, Chongjing Wang et al.

Dermatologic diseases impose a large and growing global burden, affecting billions and substantially reducing quality of life. While modern therapies can rapidly control acute symptoms, long-term outcomes are often limited by single-target paradigms, recurrent courses, and insufficient attention to systemic comorbidities. Traditional Chinese medicine (TCM) provides a complementary holistic approach via syndrome differentiation and individualized treatment, but practice is hindered by non-standardized knowledge, incomplete multimodal records, and poor scalability of expert reasoning. We propose DERM-3R, a resource-efficient multimodal agent framework to model TCM dermatologic diagnosis and treatment under limited data and compute. Based on real-world workflows, we reformulate decision-making into three core issues: fine-grained lesion recognition, multi-view lesion representation with specialist-level pathogenesis modeling, and holistic reasoning for syndrome differentiation and treatment planning. DERM-3R comprises three collaborative agents: DERM-Rec, DERM-Rep, and DERM-Reason, each targeting one component of this pipeline. Built on a lightweight multimodal LLM and partially fine-tuned on 103 real-world TCM psoriasis cases, DERM-3R performs strongly across dermatologic reasoning tasks. Evaluations using automatic metrics, LLM-as-a-judge, and physician assessment show that despite minimal data and parameter updates, DERM-3R matches or surpasses large general-purpose multimodal models. These results suggest structured, domain-aware multi-agent modeling can be a practical alternative to brute-force scaling for complex clinical tasks in dermatology and integrative medicine.

Jiajun Zhang, Jianke Zhang, Zeyu Cui et al.

Recent Large Language Models (LLMs) have demonstrated remarkable proficiency in code generation. However, their ability to create complex visualizations for scaled and structured data remains largely unevaluated and underdeveloped. To address this gap, we introduce PlotCraft, a new benchmark featuring 1k challenging visualization tasks that cover a wide range of topics, such as finance, scientific research, and sociology. The benchmark is structured around seven high-level visualization tasks and encompasses 48 distinct chart types. Crucially, it is the first to systematically evaluate both single-turn generation and multi-turn refinement across a diverse spectrum of task complexities. Our comprehensive evaluation of 23 leading LLMs on PlotCraft reveals obvious performance deficiencies in handling sophisticated visualization tasks. To bridge this performance gap, we develope SynthVis-30K, a large-scale, high-quality dataset of complex visualization code synthesized via a collaborative agent framework. Building upon this dataset, we develope PlotCraftor, a novel code generation model that achieves strong capabilities in complex data visualization with a remarkably small size. Across VisEval, PandasPlotBench, and our proposed PlotCraft, PlotCraftor shows performance comparable to that of leading proprietary approaches. Especially, on hard task, Our model achieves over 50% performance improvement. We will release the benchmark, dataset, and code at https://github.com/Speakn0w/PlotCraft-Benchmark.

Yiling Tao, Shuyi Wang, Jiaxi Yang et al.

This paper reports on findings from a comparative study on the effectiveness and efficiency of federated unlearning strategies within Federated Online Learning to Rank (FOLTR), with specific attention to systematically analysing the unlearning capabilities of methods in a verifiable manner. Federated approaches to ranking of search results have recently garnered attention to address users privacy concerns. In FOLTR, privacy is safeguarded by collaboratively training ranking models across decentralized data sources, preserving individual user data while optimizing search results based on implicit feedback, such as clicks. Recent legislation introduced across numerous countries is establishing the so called "the right to be forgotten", according to which services based on machine learning models like those in FOLTR should provide capabilities that allow users to remove their own data from those used to train models. This has sparked the development of unlearning methods, along with evaluation practices to measure whether unlearning of a user data successfully occurred. Current evaluation practices are however often controversial, necessitating the use of multiple metrics for a more comprehensive assessment -- but previous proposals of unlearning methods only used single evaluation metrics. This paper addresses this limitation: our study rigorously assesses the effectiveness of unlearning strategies in managing both under-unlearning and over-unlearning scenarios using adapted, and newly proposed evaluation metrics. Thanks to our detailed analysis, we uncover the strengths and limitations of five unlearning strategies, offering valuable insights into optimizing federated unlearning to balance data privacy and system performance within FOLTR. We publicly release our code and complete results at https://github.com/Iris1026/Unlearning-for-FOLTR.git.

Lei Zhang, Yunshui Li, Jiaming Li et al.

Some recently developed code large language models (Code LLMs) have been pre-trained on repository-level code data (Repo-Code LLMs), enabling these models to recognize repository structures and utilize cross-file information for code completion. However, in real-world development scenarios, simply concatenating the entire code repository often exceeds the context window limits of these Repo-Code LLMs, leading to significant performance degradation. In this study, we conducted extensive preliminary experiments and analyses on six Repo-Code LLMs. The results indicate that maintaining the topological dependencies of files and increasing the code file content in the completion prompts can improve completion accuracy; pruning the specific implementations of functions in all dependent files does not significantly reduce the accuracy of completions. Based on these findings, we proposed a strategy named Hierarchical Context Pruning (HCP) to construct completion prompts with high informational code content. The HCP models the code repository at the function level, maintaining the topological dependencies between code files while removing a large amount of irrelevant code content, significantly reduces the input length for repository-level code completion. We applied the HCP strategy in experiments with six Repo-Code LLMs, and the results demonstrate that our proposed method can significantly enhance completion accuracy while substantially reducing the length of input. Our code and data are available at https://github.com/Hambaobao/HCP-Coder.

Chanyong Luo, Jirui Dai, Zhendong Wang et al.

Medicine is an empirical discipline refined through long-term observation and the messy, high-variance reality of clinical practice. Physicians build diagnostic and therapeutic competence through repeated cycles of application, reflection, and improvement, forming individualized methodologies. Yet outcomes vary widely, and master physicians' knowledge systems are slow to develop and hard to transmit at scale, contributing to the scarcity of high-quality clinical expertise. To address this, we propose Med-Shicheng, a general framework that enables large language models to systematically learn and transfer distinguished physicians' diagnostic-and-therapeutic philosophy and case-dependent adaptation rules in a standardized way. Built on Tianyi, Med-Shicheng consists of five stages. We target five National Masters of Chinese Medicine or distinguished TCM physicians, curate multi-source materials, and train a single model to internalize all five knowledge systems across seven tasks, including etiology-pathogenesis analysis, syndrome diagnosis, treatment principle selection, prescription generation, prescription explanation, symptom evolution with regimen adjustment, and clinical advice. Implemented on Qwen2.5-1.5B-Base, Med-Shicheng runs on resource-constrained GPUs while achieving performance comparable to DeepSeek-R1 and GPT-5. We also examine the reliability of LLM-as-a-judge versus physician evaluation: automated judging tracks overall trends but shows bias on fine-grained individualized distinctions, highlighting the need for physician involvement when ground truth is unavailable and for domain-adapted judge models.

Yunshui Li, Binyuan Hui, Xiaobo Xia et al. · tsinghua

Contemporary practices in instruction tuning often hinge on enlarging data scaling without a clear strategy for ensuring data quality, inadvertently introducing noise that may compromise model performance. To address this challenge, we introduce \textsc{Nuggets}, a novel and efficient methodology that leverages one-shot learning to discern and select high-quality instruction data from extensive datasets. \textsc{Nuggets} assesses the potential of individual instruction examples to act as effective one-shot learning instances, thereby identifying those that can significantly improve performance across diverse tasks. \textsc{Nuggets} utilizes a scoring system based on the impact of candidate examples on the perplexity of a diverse anchor set, facilitating the selection of the most advantageous data for instruction tuning. Through comprehensive evaluations on two benchmarks, including MT-Bench and Alpaca-Eval, we show that instruction tuning with the top 1\% of examples curated by \textsc{Nuggets} substantially outperforms conventional methods employing the entire dataset.

Chengpeng Li, Mingfeng Xue, Zhenru Zhang et al.

Large reasoning models (LRMs) like OpenAI-o1 and DeepSeek-R1 have demonstrated remarkable capabilities in complex reasoning tasks through the utilization of long Chain-of-thought (CoT). However, these models often suffer from hallucinations and inefficiencies due to their reliance solely on internal reasoning processes. In this paper, we introduce START (Self-Taught Reasoner with Tools), a novel tool-integrated long CoT reasoning LLM that significantly enhances reasoning capabilities by leveraging external tools. Through code execution, START is capable of performing complex computations, self-checking, exploring diverse methods, and self-debugging, thereby addressing the limitations of LRMs. The core innovation of START lies in its self-learning framework, which comprises two key techniques: 1) Hint-infer: We demonstrate that inserting artificially designed hints (e.g., ``Wait, maybe using Python here is a good idea.'') during the inference process of a LRM effectively stimulates its ability to utilize external tools without the need for any demonstration data. Hint-infer can also serve as a simple and effective sequential test-time scaling method; 2) Hint Rejection Sampling Fine-Tuning (Hint-RFT): Hint-RFT combines Hint-infer and RFT by scoring, filtering, and modifying the reasoning trajectories with tool invocation generated by a LRM via Hint-infer, followed by fine-tuning the LRM. Through this framework, we have fine-tuned the QwQ-32B model to achieve START. On PhD-level science QA (GPQA), competition-level math benchmarks (AMC23, AIME24, AIME25), and the competition-level code benchmark (LiveCodeBench), START achieves accuracy rates of 63.6%, 95.0%, 66.7%, 47.1%, and 47.3%, respectively. It significantly outperforms the base QwQ-32B and achieves performance comparable to the state-of-the-art open-weight model R1-Distill-Qwen-32B and the proprietary model o1-Preview.

Mouxiang Chen, Binyuan Hui, Zeyu Cui et al.

It is commonly believed that scaling language models should commit a significant space or time cost, by increasing the parameters (parameter scaling) or output tokens (inference-time scaling). We introduce the third and more inference-efficient scaling paradigm: increasing the model's parallel computation during both training and inference time. We apply $P$ diverse and learnable transformations to the input, execute forward passes of the model in parallel, and dynamically aggregate the $P$ outputs. This method, namely parallel scaling (ParScale), scales parallel computation by reusing existing parameters and can be applied to any model structure, optimization procedure, data, or task. We theoretically propose a new scaling law and validate it through large-scale pre-training, which shows that a model with $P$ parallel streams is similar to scaling the parameters by $O(\log P)$ while showing superior inference efficiency. For example, ParScale can use up to 22$\times$ less memory increase and 6$\times$ less latency increase compared to parameter scaling that achieves the same performance improvement. It can also recycle an off-the-shelf pre-trained model into a parallelly scaled one by post-training on a small amount of tokens, further reducing the training budget. The new scaling law we discovered potentially facilitates the deployment of more powerful models in low-resource scenarios, and provides an alternative perspective for the role of computation in machine learning.

Jian Yang, Wei Zhang, Jiaxi Yang et al.

Recent advancement in code understanding and generation demonstrates that code LLMs fine-tuned on a high-quality instruction dataset can gain powerful capabilities to address wide-ranging code-related tasks. However, most previous existing methods mainly view each programming language in isolation and ignore the knowledge transfer among different programming languages. To bridge the gap among different programming languages, we introduce a novel multi-agent collaboration framework to enhance multilingual instruction tuning for code LLMs, where multiple language-specific intelligent agent components with generation memory work together to transfer knowledge from one language to another efficiently and effectively. Specifically, we first generate the language-specific instruction data from the code snippets and then provide the generated data as the seed data for language-specific agents. Multiple language-specific agents discuss and collaborate to formulate a new instruction and its corresponding solution (A new programming language or existing programming language), To further encourage the cross-lingual transfer, each agent stores its generation history as memory and then summarizes its merits and faults. Finally, the high-quality multilingual instruction data is used to encourage knowledge transfer among different programming languages to train Qwen2.5-xCoder. Experimental results on multilingual programming benchmarks demonstrate the superior performance of Qwen2.5-xCoder in sharing common knowledge, highlighting its potential to reduce the cross-lingual gap.

Jiajun Zhang, Zeyu Cui, Jiaxi Yang et al.

The dominant Fill-in-the-Middle (FIM) paradigm for code completion is constrained by its rigid inability to correct contextual errors and reliance on unaligned, insecure Base models. While Chat LLMs offer safety and Agentic workflows provide flexibility, they suffer from performance degradation and prohibitive latency, respectively. To resolve this dilemma, we propose Search-and-Replace Infilling (SRI), a framework that internalizes the agentic verification-and-editing mechanism into a unified, single-pass inference process. By structurally grounding edits via an explicit search phase, SRI harmonizes completion tasks with the instruction-following priors of Chat LLMs, extending the paradigm from static infilling to dynamic context-aware editing. We synthesize a high-quality dataset, SRI-200K, and fine-tune the SRI-Coder series. Extensive evaluations demonstrate that with minimal data (20k samples), SRI-Coder enables Chat models to surpass the completion performance of their Base counterparts. Crucially, unlike FIM-style tuning, SRI preserves general coding competencies and maintains inference latency comparable to standard FIM. We empower the entire Qwen3-Coder series with SRI, encouraging the developer community to leverage this framework for advanced auto-completion and assisted development.

Wei Zhang, Jian Yang, Jiaxi Yang et al.

Code large language models (LLMs) enhance programming by understanding and generating code across languages, offering intelligent feedback, bug detection, and code updates through reflection, improving development efficiency and accessibility. While benchmarks (e.g. HumanEval/LiveCodeBench) evaluate code generation and real-world relevance, previous works ignore the scenario of modifying code in repositories. Considering challenges remaining in improving reflection capabilities and avoiding data contamination in dynamic benchmarks, we introduce LiveRepoReflection, a challenging benchmark for evaluating code understanding and generation in multi-file repository contexts, featuring 1,888 rigorously filtered test cases across $6$ programming languages to ensure diversity, correctness, and high difficulty. Further, we create RepoReflection-Instruct, a large-scale, quality-filtered instruction-tuning dataset derived from diverse sources, used to train RepoReflectionCoder through a two-turn dialogue process involving code generation and error-driven repair. The leaderboard evaluates over 40 LLMs to reflect the model performance of repository-based code reflection.

Zhi Liu, Tao Yang, Jing Wang et al.

Natural medicines, particularly Traditional Chinese Medicine (TCM), are gaining global recognition for their therapeutic potential in addressing human symptoms and diseases. TCM, with its systematic theories and extensive practical experience, provides abundant resources for healthcare. However, the effective application of TCM requires precise syndrome diagnosis, determination of treatment principles, and prescription formulation, which demand decades of clinical expertise. Despite advancements in TCM-based decision systems, machine learning, and deep learning research, limitations in data and single-objective constraints hinder their practical application. In recent years, large language models (LLMs) have demonstrated potential in complex tasks, but lack specialization in TCM and face significant challenges, such as too big model scale to deploy and issues with hallucination. To address these challenges, we introduce Tianyi with 7.6-billion-parameter LLM, a model scale proper and specifically designed for TCM, pre-trained and fine-tuned on diverse TCM corpora, including classical texts, expert treatises, clinical records, and knowledge graphs. Tianyi is designed to assimilate interconnected and systematic TCM knowledge through a progressive learning manner. Additionally, we establish TCMEval, a comprehensive evaluation benchmark, to assess LLMs in TCM examinations, clinical tasks, domain-specific question-answering, and real-world trials. The extensive evaluations demonstrate the significant potential of Tianyi as an AI assistant in TCM clinical practice and research, bridging the gap between TCM knowledge and practical application.

Wendi Wang, Jiaxi Yang, Yongkang Du et al.

Graph diffusion models have gained significant attention in graph generation tasks, but they often inherit and amplify topology biases from sensitive attributes (e.g. gender, age, region), leading to unfair synthetic graphs. Existing fair graph generation using diffusion models is limited to specific graph-based applications with complete labels or requires simultaneous updates for graph structure and node attributes, making them unsuitable for general usage. To relax these limitations by applying the debiasing method directly on graph topology, we propose Fair Graph Diffusion Model (FairGDiff), a counterfactual-based one-step solution that mitigates topology biases while balancing fairness and utility. In detail, we construct a causal model to capture the relationship between sensitive attributes, biased link formation, and the generated graph structure. By answering the counterfactual question "Would the graph structure change if the sensitive attribute were different?", we estimate an unbiased treatment and incorporate it into the diffusion process. FairGDiff integrates counterfactual learning into both forward diffusion and backward denoising, ensuring that the generated graphs are independent of sensitive attributes while preserving structural integrity. Extensive experiments on real-world datasets demonstrate that FairGDiff achieves a superior trade-off between fairness and utility, outperforming existing fair graph generation methods while maintaining scalability.

Jiaxi Yang, Mengqi Zhang, Yiqiao Jin et al. · amazon-science

Large Language Model-based Multi-Agent Systems (MASs) have emerged as a powerful paradigm for tackling complex tasks through collaborative intelligence. However, the topology of these systems--how agents in MASs should be configured, connected, and coordinated--remains largely unexplored. In this position paper, we call for a paradigm shift toward \emph{topology-aware MASs} that explicitly model and dynamically optimize the structure of inter-agent interactions. We identify three fundamental components--agents, communication links, and overall topology--that collectively determine the system's adaptability, efficiency, robustness, and fairness. To operationalize this vision, we introduce a systematic three-stage framework: 1) agent selection, 2) structure profiling, and 3) topology synthesis. This framework not only provides a principled foundation for designing MASs but also opens new research frontiers across language modeling, reinforcement learning, graph learning, and generative modeling to ultimately unleash their full potential in complex real-world applications. We conclude by outlining key challenges and opportunities in MASs evaluation. We hope our framework and perspectives offer critical new insights in the era of agentic AI.

Yurui Dong, Luozhijie Jin, Yao Yang et al.

Purpose: Emotion is a fundamental component of human communication, shaping understanding, trust, and engagement across domains such as education, healthcare, and mental health. While large language models (LLMs) exhibit strong reasoning and knowledge generation capabilities, they still struggle to express emotions in a consistent, controllable, and contextually appropriate manner. This limitation restricts their potential for authentic human-AI interaction. Methods: We propose a controllable emotion generation framework based on Emotion Vectors (EVs) - latent representations derived from internal activation shifts between neutral and emotion-conditioned responses. By injecting these vectors into the hidden states of pretrained LLMs during inference, our method enables fine-grained, continuous modulation of emotional tone without any additional training or architectural modification. We further provide theoretical analysis proving that EV steering enhances emotional expressivity while maintaining semantic fidelity and linguistic fluency. Results: Extensive experiments across multiple LLM families show that the proposed approach achieves consistent emotional alignment, stable topic adherence, and controllable affect intensity. Compared with existing prompt-based and fine-tuning-based baselines, our method demonstrates superior flexibility and generalizability. Conclusion: Emotion Vector (EV) steering provides an efficient and interpretable means of bridging rational reasoning and affective understanding in large language models, offering a promising direction for building emotionally resonant AI systems capable of more natural human-machine interaction.

Jiaxi Yang, Haowen Hou

Existing Visual Language Modelsoften struggle with information loss and limited reasoning abilities when handling high-resolution web interfaces that combine complex visual, textual, and interactive elements. These challenges are particularly evident in tasks requiring webpage layout comprehension and multi-step interactive reasoning. To address these challenges, we propose RWKV-UI, a Visual Language Model based on the RWKV architecture, specifically designed to handle high-resolution UI images. During model training, we introduce layout detection as a visual prompt to help the model better understand the webpage layout structures. Additionally, we design a visual prompt based on the Chain-of-Thought(CoT) mechanism, which enhances the model's ability to understand and reason about webpage content through reasoning chains. Experimental results show that RWKV-UI demonstrates significant performance improvements in high-resolution UI understanding and interactive reasoning tasks.

Jiaxi Yang, Binyuan Hui, Min Yang et al.

Despite the advancements in in-context learning (ICL) for large language models (LLMs), current research centers on specific prompt engineering, such as demonstration selection, with the expectation that a single iteration of demonstrations processing can generalize effectively to a given test sample. However, this perspective overlooks the potential benefits derived from multiple iterations involving demonstrations, a practice aligning more closely with the iterative decision-making process exhibited by humans, who often learn through analogy. In this study, we introduce a novel two-stage framework to boost ICL in LLMs. Specifically, our framework delineates the ICL process into two distinct stages: Deep-Thinking and test stages. The Deep-Thinking stage incorporates a unique attention mechanism, i.e., iterative enhanced attention, which enables multiple rounds of information accumulation. This mechanism operates by manipulating the Key-Value matrices without training, fostering enhanced understanding capabilities in LLMs by thinking demonstrations multiple times. We evaluated Deep-Thinking across a range of benchmarks and LLMs, showing its superior performance over vanilla ICL methods and its effectiveness in challenging tasks where demonstration selection is infeasible.

Jinyang Li, Binyuan Hui, Ge Qu et al.

Text-to-SQL parsing, which aims at converting natural language instructions into executable SQLs, has gained increasing attention in recent years. In particular, Codex and ChatGPT have shown impressive results in this task. However, most of the prevalent benchmarks, i.e., Spider, and WikiSQL, focus on database schema with few rows of database contents leaving the gap between academic study and real-world applications. To mitigate this gap, we present Bird, a big benchmark for large-scale database grounded in text-to-SQL tasks, containing 12,751 pairs of text-to-SQL data and 95 databases with a total size of 33.4 GB, spanning 37 professional domains. Our emphasis on database values highlights the new challenges of dirty database contents, external knowledge between NL questions and database contents, and SQL efficiency, particularly in the context of massive databases. To solve these problems, text-to-SQL models must feature database value comprehension in addition to semantic parsing. The experimental results demonstrate the significance of database values in generating accurate text-to-SQLs for big databases. Furthermore, even the most effective text-to-SQL models, i.e. ChatGPT, only achieves 40.08% in execution accuracy, which is still far from the human result of 92.96%, proving that challenges still stand. Besides, we also provide an efficiency analysis to offer insights into generating text-to-efficient-SQLs that are beneficial to industries. We believe that BIRD will contribute to advancing real-world applications of text-to-SQL research. The leaderboard and source code are available: https://bird-bench.github.io/.

Kehan Wang, Jiaxi Yang, Hongjun Wu

While in real life everyone behaves themselves at least to some extent, it is much more difficult to expect people to behave themselves on the internet, because there are few checks or consequences for posting something toxic to others. Yet, for people on the other side, toxic texts often lead to serious psychological consequences. Detecting such toxic texts is challenging. In this paper, we attempt to build a toxicity detector using machine learning methods including CNN, Naive Bayes model, as well as LSTM. While there has been numerous groundwork laid by others, we aim to build models that provide higher accuracy than the predecessors. We produced very high accuracy models using LSTM and CNN, and compared them to the go-to solutions in language processing, the Naive Bayes model. A word embedding approach is also applied to empower the accuracy of our models.

Lei Chen, Fajie Yuan, Jiaxi Yang et al.

Sequential recommender systems (SRS) have become a research hotspot due to its power in modeling user dynamic interests and sequential behavioral patterns. To maximize model expressive ability, a default choice is to apply a larger and deeper network architecture, which, however, often brings high network latency when generating online recommendations. Naturally, we argue that compressing the heavy recommendation models into middle- or light- weight neural networks is of great importance for practical production systems. To realize such a goal, we propose AdaRec, a knowledge distillation (KD) framework which compresses knowledge of a teacher model into a student model adaptively according to its recommendation scene by using differentiable Neural Architecture Search (NAS). Specifically, we introduce a target-oriented distillation loss to guide the structure search process for finding the student network architecture, and a cost-sensitive loss as constraints for model size, which achieves a superior trade-off between recommendation effectiveness and efficiency. In addition, we leverage Earth Mover's Distance (EMD) to realize many-to-many layer mapping during knowledge distillation, which enables each intermediate student layer to learn from other intermediate teacher layers adaptively. Extensive experiments on real-world recommendation datasets demonstrate that our model achieves competitive or better accuracy with notable inference speedup comparing to strong counterparts, while discovering diverse neural architectures for sequential recommender models under different recommendation scenes.

Lei Chen, Fajie Yuan, Jiaxi Yang et al.

Making accurate recommendations for cold-start users has been a longstanding and critical challenge for recommender systems (RS). Cross-domain recommendations (CDR) offer a solution to tackle such a cold-start problem when there is no sufficient data for the users who have rarely used the system. An effective approach in CDR is to leverage the knowledge (e.g., user representations) learned from a related but different domain and transfer it to the target domain. Fine-tuning works as an effective transfer learning technique for this objective, which adapts the parameters of a pre-trained model from the source domain to the target domain. However, current methods are mainly based on the global fine-tuning strategy: the decision of which layers of the pre-trained model to freeze or fine-tune is taken for all users in the target domain. In this paper, we argue that users in RS are personalized and should have their own fine-tuning policies for better preference transfer learning. As such, we propose a novel User-specific Adaptive Fine-tuning method (UAF), selecting which layers of the pre-trained network to fine-tune, on a per-user basis. Specifically, we devise a policy network with three alternative strategies to automatically decide which layers to be fine-tuned and which layers to have their parameters frozen for each user. Extensive experiments show that the proposed UAF exhibits significantly better and more robust performance for user cold-start recommendation.