Dingjie Song

Huang Huang, Fei Yu, Jianqing Zhu et al.

This paper is devoted to the development of a localized Large Language Model (LLM) specifically for Arabic, a language imbued with unique cultural characteristics inadequately addressed by current mainstream models. Significant concerns emerge when addressing cultural sensitivity and local values. To address this, the paper proposes a comprehensive solution that includes further pre-training with Arabic texts, Supervised Fine-Tuning (SFT) utilizing native Arabic instructions, and GPT-4 responses in Arabic, alongside Reinforcement Learning with AI Feedback (RLAIF) employing a reward model attuned to local culture and values. The goal is to cultivate culturally cognizant and value-aligned Arabic LLMs capable of accommodating the diverse, application-specific needs of Arabic-speaking communities. Comprehensive evaluations reveal that the resulting model, dubbed `AceGPT', sets the state-of-the-art standard for open Arabic LLMs across various benchmarks. Codes, data, and models are in https://github.com/FreedomIntelligence/AceGPT.

Xidong Wang, Guiming Hardy Chen, Dingjie Song et al.

Large Language Models (LLMs) provide a possibility to make a great breakthrough in medicine. The establishment of a standardized medical benchmark becomes a fundamental cornerstone to measure progression. However, medical environments in different regions have their local characteristics, e.g., the ubiquity and significance of traditional Chinese medicine within China. Therefore, merely translating English-based medical evaluation may result in \textit{contextual incongruities} to a local region. To solve the issue, we propose a localized medical benchmark called CMB, a Comprehensive Medical Benchmark in Chinese, designed and rooted entirely within the native Chinese linguistic and cultural framework. While traditional Chinese medicine is integral to this evaluation, it does not constitute its entirety. Using this benchmark, we have evaluated several prominent large-scale LLMs, including ChatGPT, GPT-4, dedicated Chinese LLMs, and LLMs specialized in the medical domain. We hope this benchmark provide first-hand experience in existing LLMs for medicine and also facilitate the widespread adoption and enhancement of medical LLMs within China. Our data and code are publicly available at https://github.com/FreedomIntelligence/CMB.

Dingjie Song, Tianlong Xu, Yi-Fan Zhang et al.

Assessing student handwritten scratchwork is crucial for personalized educational feedback but presents unique challenges due to diverse handwriting, complex layouts, and varied problem-solving approaches. Existing educational NLP primarily focuses on textual responses and neglects the complexity and multimodality inherent in authentic handwritten scratchwork. Current multimodal large language models (MLLMs) excel at visual reasoning but typically adopt an "examinee perspective", prioritizing generating correct answers rather than diagnosing student errors. To bridge these gaps, we introduce ScratchMath, a novel benchmark specifically designed for explaining and classifying errors in authentic handwritten mathematics scratchwork. Our dataset comprises 1,720 mathematics samples from Chinese primary and middle school students, supporting two key tasks: Error Cause Explanation (ECE) and Error Cause Classification (ECC), with seven defined error types. The dataset is meticulously annotated through rigorous human-machine collaborative approaches involving multiple stages of expert labeling, review, and verification. We systematically evaluate 16 leading MLLMs on ScratchMath, revealing significant performance gaps relative to human experts, especially in visual recognition and logical reasoning. Proprietary models notably outperform open-source models, with large reasoning models showing strong potential for error explanation. All evaluation data and frameworks are publicly available to facilitate further research.

Zhenyang Cai, Junying Chen, Rongsheng Wang et al.

Medical imaging provides essential visual insights for diagnosis, and multimodal large language models (MLLMs) are increasingly utilized for its analysis due to their strong generalization capabilities; however, the underlying factors driving this generalization remain unclear. Current research suggests that multi-task training outperforms single-task as different tasks can benefit each other, but they often overlook the internal relationships within these tasks. To analyze this phenomenon, we attempted to employ compositional generalization (CG), which refers to the models' ability to understand novel combinations by recombining learned elements, as a guiding framework. Since medical images can be precisely defined by Modality, Anatomical area, and Task, naturally providing an environment for exploring CG, we assembled 106 medical datasets to create Med-MAT for comprehensive experiments. The experiments confirmed that MLLMs can use CG to understand unseen medical images and identified CG as one of the main drivers of the generalization observed in multi-task training. Additionally, further studies demonstrated that CG effectively supports datasets with limited data and confirmed that MLLMs can achieve CG across classification and detection tasks, underscoring its broader generalization potential. Med-MAT is available at https://github.com/FreedomIntelligence/Med-MAT.

Zhiling Yan, Dingjie Song, Hanrong Zhang et al.

Self-evolving agents requires adaptation after deployment, but existing approaches assume a usable learning loop, such as curated skills, successful trajectories, or verifier signals. Real open-world deployments may provide none of these, offering only a task prompt. In this work, we study open-world self-evolution, where an agent must build both its skills and its own verification signals from scratch, using open-world resources but no target-task supervision. We propose OpenSkill, a framework that bootstraps this loop: it acquires grounded knowledge and verification anchors from documentation, repositories, and the web, synthesizes them into transferable skills, and refines those skills against self-built virtual tasks grounded in the anchors rather than in target answers. The open world thus supplies both the knowledge to be learned and a supervision-independent practice environment, with target-task supervision reserved for final evaluation. Across three benchmarks and two target agents, OpenSkill attains the best automated pass rate while satisfying the no-supervision constraint. Analysis shows its skills transfer across models without model-specific adaptation, and its self-built verifier aligns with ground-truth outcomes despite never accessing them.

Yi-Fan Zhang, Hang Li, Dingjie Song et al.

Large Language Models (LLMs), such as GPT-4, have demonstrated impressive mathematical reasoning capabilities, achieving near-perfect performance on benchmarks like GSM8K. However, their application in personalized education remains limited due to an overemphasis on correctness over error diagnosis and feedback generation. Current models fail to provide meaningful insights into the causes of student mistakes, limiting their utility in educational contexts. To address these challenges, we present three key contributions. First, we introduce \textbf{MathCCS} (Mathematical Classification and Constructive Suggestions), a multi-modal benchmark designed for systematic error analysis and tailored feedback. MathCCS includes real-world problems, expert-annotated error categories, and longitudinal student data. Evaluations of state-of-the-art models, including \textit{Qwen2-VL}, \textit{LLaVA-OV}, \textit{Claude-3.5-Sonnet} and \textit{GPT-4o}, reveal that none achieved classification accuracy above 30\% or generated high-quality suggestions (average scores below 4/10), highlighting a significant gap from human-level performance. Second, we develop a sequential error analysis framework that leverages historical data to track trends and improve diagnostic precision. Finally, we propose a multi-agent collaborative framework that combines a Time Series Agent for historical analysis and an MLLM Agent for real-time refinement, enhancing error classification and feedback generation. Together, these contributions provide a robust platform for advancing personalized education, bridging the gap between current AI capabilities and the demands of real-world teaching.

Xianghong Fang, Litao Guo, Hengchao Chen et al.

The success of autoregressive models largely depends on the effectiveness of vector quantization, a technique that discretizes continuous features by mapping them to the nearest code vectors within a learnable codebook. Two critical issues in existing vector quantization methods are training instability and codebook collapse. Training instability arises from the gradient discrepancy introduced by the straight-through estimator, especially in the presence of significant quantization errors, while codebook collapse occurs when only a small subset of code vectors are utilized during training. A closer examination of these issues reveals that they are primarily driven by a mismatch between the distributions of the features and code vectors, leading to unrepresentative code vectors and significant data information loss during compression. To address this, we employ the Wasserstein distance to align these two distributions, achieving near 100\% codebook utilization and significantly reducing the quantization error. Both empirical and theoretical analyses validate the effectiveness of the proposed approach.

Dingjie Song, Sicheng Lai, Mingxuan Wang et al.

The rapid advancement of multimodal large language models (MLLMs) has significantly enhanced performance across benchmarks. However, data contamination-unintentional memorization of benchmark data during model training-poses critical challenges for fair evaluation. Existing detection methods for unimodal large language models (LLMs) are inadequate for MLLMs due to multimodal data complexity and multi-phase training. We systematically analyze multimodal data contamination using our analytical framework, MM-Detect, which defines two contamination categories-unimodal and cross-modal-and effectively quantifies contamination severity across multiple-choice and caption-based Visual Question Answering tasks. Evaluations on twelve MLLMs and five benchmarks reveal significant contamination, particularly in proprietary models and older benchmarks. Crucially, contamination sometimes originates during unimodal pre-training rather than solely from multimodal fine-tuning. Our insights refine contamination understanding, guiding evaluation practices and improving multimodal model reliability.

Ran Wang, Kun Tao, Dingjie Song et al.

Existing question answering systems can only predict answers without explicit reasoning processes, which hinder their explainability and make us overestimate their ability of understanding and reasoning over natural language. In this work, we propose a novel task of reading comprehension, in which a model is required to provide final answers and reasoning processes. To this end, we introduce a formalism for reasoning over unstructured text, namely Text Reasoning Meaning Representation (TRMR). TRMR consists of three phrases, which is expressive enough to characterize the reasoning process to answer reading comprehension questions. We develop an annotation platform to facilitate TRMR's annotation, and release the R3 dataset, a \textbf{R}eading comprehension benchmark \textbf{R}equiring \textbf{R}easoning processes. R3 contains over 60K pairs of question-answer pairs and their TRMRs. Our dataset is available at: \url{http://anonymous}.