We introduce Baichuan-M3, a medical-enhanced large language model engineered to shift the paradigm from passive question-answering to active, clinical-grade decision support. Addressing the limitations of existing systems in open-ended consultations, Baichuan-M3 utilizes a specialized training pipeline to model the systematic workflow of a physician. Key capabilities include: (i) proactive information acquisition to resolve ambiguity; (ii) long-horizon reasoning that unifies scattered evidence into coherent diagnoses; and (iii) adaptive hallucination suppression to ensure factual reliability. Empirical evaluations demonstrate that Baichuan-M3 achieves state-of-the-art results on HealthBench, the newly introduced HealthBench-Hallu and ScanBench, significantly outperforming GPT-5.2 in clinical inquiry, advisory and safety. The models are publicly available at https://huggingface.co/collections/baichuan-inc/baichuan-m3.

Accurately identifying the diffusion source is an effective control strategy in complex network diffusion processes. Current methods predominantly focus on temporal and directional characteristics of observed nodes, yet frequently overlook structural properties and second-order neighborhood influences. To address this, we propose an Integrated Dual-Mapping Twin-Graph Neural Network (IDM-TGN) framework for source identification. IDM-TGN consists of two core modules: Integrated Structural Dual-Mapping (ISDM) module and Twin-GNN Fusion Architecture (TG-FA). The ISDM module leverages the structural holes centered on observation nodes to extract and integrate multiple distinct diffusion features into a unified feature matrix. The TG-FA module employs two distinct graph neural networks to directly identify the diffusion source from the input feature matrix. Extensive experiments on diverse synthetic and real networks demonstrate the effectiveness and feasibility of IDM-TGN for source identification.

With the increasing availability of multimodal remote sensing (RS) data, semantic segmentation that leverages complementary information from true orthophotos (TOP) and digital surface models (DSM) has become essential for urban analysis. Diffusion-based segmentation provides an effective iterative refinement mechanism for modeling complex multimodal distributions; however, conventional pixel-wise supervision emphasizes local accuracy while overlooking global distribution alignment, often leading to inconsistent predictions and blurred object boundaries. Although maximum mean discrepancy (MMD) measures global statistical differences between predicted and ground-truth distributions, its effectiveness in high-dimensional class-probability spaces is limited by directional cancellation effects that reduce sensitivity to complex distribution shifts. To address this issue, we propose a projection-kernel regularized diffusion-based multimodal RS segmentation framework that enforces global statistical alignment through distribution-level regularization rather than modifying the intrinsic diffusion process. The proposed regularization performs multi-directional projections of high-dimensional class-probability vectors onto one-dimensional subspaces and derives a closed-form kernel integration to avoid numerical sampling across projection directions, enabling efficient and stable global distribution matching. In addition, a Cross-Attention Dual-Encoder Fusion (CADEF) module is introduced to alleviate geometry-texture misalignment, and a Hierarchical EMA-Gated Recursive Denoising (HERD) mechanism is designed to stabilize multiscale feature refinement. Experiments on the ISPRS Vaihingen and Potsdam benchmarks demonstrate that the proposed regularization consistently improves segmentation accuracy over state-of-the-art CNN-, Transformer-, and diffusion-based baselines, yielding enhanced global consistency and sharper boundary delineation. Code is available at: https://github.com/tonyy127/PKDiff

Baichuan-M4 is Baichuan Intelligence's clinical-grade medical large model, designed for continuous care rather than single-turn medical question answering. It is built as a coordinated medical agent system around three pillars: Baichuan-Harness, a unified runtime that keeps reinforcement-learning training and real-world deployment consistent while enforcing action constraints, tool use, long-term patient memory, and multi-agent coordination; a core reasoning model trained with a continuous-care reinforcement-learning framework that integrates span-level reward modeling (SPAR++), reasoning-path compression, curriculum learning, and stabilized policy optimization; and a clinical tool layer for patient-memory management, authoritative evidence-based retrieval, and multimodal medical perception across documents, X-rays, and dermatology. On a cross-dimensional medical evaluation suite, Baichuan-M4 attains leading results in static medical knowledge and safety, dynamic OSCE-style consultation, long-context clinical memory, evidence-based retrieval, medical document OCR, and multimodal image understanding, while lowering the hallucination rate to 3.3

As large language models (LLMs) advance in conversational and reasoning capabilities, their practical application in healthcare has become a critical research focus. However, there is a notable gap between the performance of medical LLMs on static benchmarks such as USMLE and their utility in real-world clinical decision-making. This discrepancy arises because traditional exams fail to capture the dynamic, interactive nature of medical consultations. To address this challenge, we introduce a novel dynamic verification framework that moves beyond static answer verifier, establishing a large-scale, high-fidelity interactive reinforcement learning system. Our framework comprises two key components: a Patient Simulator that creates realistic clinical environments using de-identified medical records, and a Clinical Rubrics Generator that dynamically produces multi-dimensional evaluation metrics. Building on this foundation, we develop Baichuan-M2, a 32B-parameter medical augmented reasoning model trained through a multi-stage reinforcement learning strategy with an improved Group Relative Policy Optimization (GRPO) algorithm. Evaluated on HealthBench, Baichuan-M2 outperforms all other open-source models and most advanced closed-source counterparts, achieving a score above 32 on the challenging HealthBench Hard benchmark-previously exceeded only by GPT-5. Our work demonstrates that robust dynamic verifier system is essential for aligning LLM capabilities with practical clinical applications, establishing a new Pareto front in the performance-parameter trade-off for medical AI deployment.

We introduce Baichuan-Omni-1.5, an omni-modal model that not only has omni-modal understanding capabilities but also provides end-to-end audio generation capabilities. To achieve fluent and high-quality interaction across modalities without compromising the capabilities of any modality, we prioritized optimizing three key aspects. First, we establish a comprehensive data cleaning and synthesis pipeline for multimodal data, obtaining about 500B high-quality data (text, audio, and vision). Second, an audio-tokenizer (Baichuan-Audio-Tokenizer) has been designed to capture both semantic and acoustic information from audio, enabling seamless integration and enhanced compatibility with MLLM. Lastly, we designed a multi-stage training strategy that progressively integrates multimodal alignment and multitask fine-tuning, ensuring effective synergy across all modalities. Baichuan-Omni-1.5 leads contemporary models (including GPT4o-mini and MiniCPM-o 2.6) in terms of comprehensive omni-modal capabilities. Notably, it achieves results comparable to leading models such as Qwen2-VL-72B across various multimodal medical benchmarks.

According to the Test-Time Scaling, the integration of External Slow-Thinking with the Verify mechanism has been demonstrated to enhance multi-round reasoning in large language models (LLMs). However, in the multimodal (MM) domain, there is still a lack of a strong MM-Verifier. In this paper, we introduce MM-Verifier and MM-Reasoner to enhance multimodal reasoning through longer inference and more robust verification. First, we propose a two-step MM verification data synthesis method, which combines a simulation-based tree search with verification and uses rejection sampling to generate high-quality Chain-of-Thought (COT) data. This data is then used to fine-tune the verification model, MM-Verifier. Additionally, we present a more efficient method for synthesizing MMCOT data, bridging the gap between text-based and multimodal reasoning. The synthesized data is used to fine-tune MM-Reasoner. Our MM-Verifier outperforms all larger models on the MathCheck, MathVista, and MathVerse benchmarks. Moreover, MM-Reasoner demonstrates strong effectiveness and scalability, with performance improving as data size increases. Finally, our approach achieves strong performance when combining MM-Reasoner and MM-Verifier, reaching an accuracy of 65.3 on MathVista, surpassing GPT-4o (63.8) with 12 rollouts.

Background Patients undergoing peritoneal dialysis (PD) frequently experience a progressive loss of lean mass and an increase in fat mass. Quantitative body composition analysis is critical for predicting mortality; In this study, we developed a deep learning model that can automatically segment abdominal CT images to quantify body composition and verified its clinical applicability. Material and Methods The deep learning model, which quantifies tissue components at the L1–L5 vertebral levels, was developed using data from outpatient and patient undergoing PD cohorts. To establish ground-truth labels, two specialists manually segmented visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT), and skeletal muscle (SM). Sarcopenia was classified using accepted SM index cutoffs and SAT- and VAT-based median values. Cox regression analysis and Kaplan–Meier curves were utilized to determine the relationship between these variables and mortality. Results The mean dice similarity coefficient values for SM, SAT, and VAT were high in the test set. To evaluate their clinical utility, we assessed the predictive capacity of body composition measurements for clinical outcomes among patients undergoing PD. Upon adjusting for age, gender, SAT, and VAT, we found an independent association between skeletal muscle index (SMI) and mortality (P < 0.05). Conclusion This study provides an automated body composition estimation approach from the L1–L5 vertebral levels. Automated SMI assessments were found to be predictive of mortality in patients undergoing PD, providing evidence for the wider implementation potential of this methodology in clinical practice.

With the rapid progress of Multimodal LLMs, evaluating their mathematical reasoning capabilities has become an increasingly important research direction. In particular, visual-textual mathematical reasoning serves as a key indicator of an MLLM's ability to comprehend and solve complex, multi-step quantitative problems. While existing benchmarks such as MathVista and MathVerse have advanced the evaluation of multimodal math proficiency, they primarily rely on digitally rendered content and fall short in capturing the complexity of real-world scenarios. To bridge this gap, we introduce MathScape, a novel benchmark focused on assessing MLLMs' reasoning ability in realistic mathematical contexts. MathScape comprises 1,369 high-quality math problems paired with human-captured real-world images, closely reflecting the challenges encountered in practical educational settings. We conduct a thorough multi-dimensional evaluation across nine leading closed-source MLLMs, three open-source MLLMs with over 20 billion parameters, and seven smaller-scale MLLMs. Our results show that even SOTA models struggle with real-world math tasks, lagging behind human performance-highlighting critical limitations in current model capabilities. Moreover, we find that strong performance on synthetic or digitally rendered images does not guarantee similar effectiveness on real-world tasks. This underscores the necessity of MathScape in the next stage of multimodal mathematical reasoning.

The adeptness of Large Language Models (LLMs) in comprehending and following natural language instructions is critical for their deployment in sophisticated real-world applications. Existing evaluations mainly focus on fragmented constraints or narrow scenarios, but they overlook the comprehensiveness and authenticity of constraints from the user’s perspective. To bridge this gap, we propose CFBench, a large-scale Chinese Comprehensive Constraints Following Benchmark for LLMs, featuring 1,000 curated samples that cover more than 200 real-life scenarios and over 50 NLP tasks. CFBench meticulously compiles constraints from real-world instructions and constructs an innovative systematic framework for constraint types, which includes 10 primary categories and over 25 subcategories, and ensures each constraint is seamlessly integrated within the instructions. To make certain that the evaluation of LLM outputs aligns with user perceptions, we propose an advanced methodology that integrates multi-dimensional assessment criteria with requirement prioritization, covering various perspectives of constraints, instructions, and requirement fulfillment. Evaluating current leading LLMs on CFBench reveals substantial room for improvement in constraints following, and we further investigate influencing factors and enhancement strategies. The data and code will be made available.

Large Language Models (LLMs) have exhibited significant potential in performing diverse tasks, including the ability to call functions or use external tools to enhance their performance. While current research on function calling by LLMs primarily focuses on single-turn interactions, this paper addresses the overlooked necessity for LLMs to engage in multi-turn function calling—critical for handling compositional, real-world queries that require planning with functions but not only use functions. To facilitate this, we introduce an approach, BUTTON, which generates synthetic compositional instruction tuning data via bottom-up instruction construction and top-down trajectory generation. In the bottom-up phase, we generate simple atomic tasks based on real-world scenarios and build compositional tasks using heuristic strategies based on atomic tasks. Corresponding function definitions are then synthesized for these compositional tasks. The top-down phase features a multi-agent environment where interactions among simulated humans, assistants, and tools are utilized to gather multi-turn function calling trajectories. This approach ensures task compositionality and allows for effective function and trajectory generation by examining atomic tasks within compositional tasks. We produce a dataset BUTTONInstruct comprising 8k data points and demonstrate its effectiveness through extensive experiments across various LLMs.

Accurate prediction of drug-target binding affinity (DTA) is pivotal for drug discovery, yet current computational methods struggle to integrate heterogeneous biomedical knowledge and capture complex molecular interactions. We present LKE-DTA, a novel deep learning framework that synergistically integrates large language models (LLMs) with knowledge graphs (KGs) to create comprehensive multi-dimensional representations for drugs and proteins. Besides, we propose a Dual Multi-Head Attention mechanism that dynamically fuses heterogeneous embeddings and captures complex dependencies, thereby significantly enhancing predictive accuracy. On benchmark datasets, comprehensive evaluations under fivefold cross-validation demonstrate that LKE-DTA consistently outperforms state-of-the-art methods. On Davis, it reduces MSE and MAE by 14.7% and 8.2%, increases CI and r by 0.9% and 3.4%. On KIBA, it achieves reductions of 4.6% in MSE and 5.3% in MAE, with improvements of 0.8% in CI and 1.5% in r, while maintaining robust convergence. In cold-start evaluation, LKE-DTA shows strong generalization: in the Cold Drug setting, CI and r improve by 2.4% and 9.6%; in the Cold Target setting, MSE, MAE, CI, and r improve by 10.2%, 12.2%, 6.6%, and 9.0%. On an independent test set, it achieves the lowest MSE and MAE and the highest CI and r, surpassing the best baseline by 9.5%, 13.0%, 6.6% and 9.6%, respectively. This work demonstrates the significant potential of combining LLMs with KGs to address biomedical challenges, opening new avenues for drug design and precision medicine research.

Large Language Models (LLMs) have shown remarkable capabilities in reasoning, exemplified by the success of OpenAI-o1 and DeepSeek-R1. However, integrating reasoning with external search processes remains challenging, especially for complex multi-hop questions requiring multiple retrieval steps. We propose ReSearch, a novel framework that trains LLMs to Reason with Search via reinforcement learning without using any supervised data on reasoning steps. Our approach treats search operations as integral components of the reasoning chain, where when and how to perform searches is guided by text-based thinking, and search results subsequently influence further reasoning. We train ReSearch on Qwen2.5-7B(-Instruct) and Qwen2.5-32B(-Instruct) models and conduct extensive experiments. Despite being trained on only one dataset, our models demonstrate strong generalizability across various benchmarks. Analysis reveals that ReSearch naturally elicits advanced reasoning capabilities such as reflection and self-correction during the reinforcement learning process.

As Large Language Models (LLMs) are pre-trained on ultra-large-scale corpora, the problem of data contamination is becoming increasingly serious, and there is a risk that static evaluation benchmarks overestimate the performance of LLMs. To address this, this paper proposes a dynamic data evaluation method called AdEval (Alignment-based Dynamic Evaluation). AdEval first extracts knowledge points and main ideas from static datasets to achieve dynamic alignment with the core content of static benchmarks, and by avoiding direct reliance on static datasets, it inherently reduces the risk of data contamination from the source. It then obtains background information through online searches to generate detailed descriptions of the knowledge points. Finally, it designs questions based on Bloom's cognitive hierarchy across six dimensions-remembering, understanding, applying, analyzing, evaluating, and creating to enable multi-level cognitive assessment. Additionally, AdEval controls the complexity of dynamically generated datasets through iterative question reconstruction. Experimental results on multiple datasets show that AdEval effectively alleviates the impact of data contamination on evaluation results, solves the problems of insufficient complexity control and single-dimensional evaluation, and improves the fairness, reliability and diversity of LLMs evaluation.

We introduce Baichuan Alignment, a detailed analysis of the alignment techniques employed in the Baichuan series of models. This represents the industry's first comprehensive account of alignment methodologies, offering valuable insights for advancing AI research. We investigate the critical components that enhance model performance during the alignment process, including optimization methods, data strategies, capability enhancements, and evaluation processes. The process spans three key stages: Prompt Augmentation System(PAS), Supervised Fine-Tuning(SFT), and Preference Alignment. The problems encountered, the solutions applied, and the improvements made are thoroughly recorded. Through comparisons across well-established benchmarks, we highlight the technological advancements enabled by Baichuan Alignment. Baichuan-Instruct is an internal model, while Qwen2-Nova-72B and Llama3-PBM-Nova-70B are instruct versions of the Qwen2-72B and Llama-3-70B base models, optimized through Baichuan Alignment. Baichuan-Instruct demonstrates significant improvements in core capabilities, with user experience gains ranging from 17 performs exceptionally well on specialized benchmarks. In open-source benchmark evaluations, both Qwen2-Nova-72B and Llama3-PBM-Nova-70B consistently outperform their respective official instruct versions across nearly all datasets. This report aims to clarify the key technologies behind the alignment process, fostering a deeper understanding within the community. Llama3-PBM-Nova-70B model is available at https://huggingface.co/PKU-Baichuan-MLSystemLab/Llama3-PBM-Nova-70B.

While in-context learning with large language models (LLMs) has shown impressive performance, we have discovered a unique miscalibration behavior where both correct and incorrect predictions are assigned the same level of confidence. We refer to this phenomenon as indiscriminate miscalibration. We found that traditional calibration metrics, such as Expected Calibrated Errors (ECEs), are unable to capture this behavior effectively. To address this issue, we propose new metrics to measure the severity of indiscriminate miscalibration. Additionally, we develop a novel in-context comparative inference method to alleviate miscalibrations and improve classification performance. Through extensive experiments on five datasets, we demonstrate that our proposed method can achieve more accurate and calibrated predictions compared to regular zero-shot and few-shot prompting.

A critical question about Large Language Models (LLMs) is whether their apparent deficiency in mathematical reasoning is inherent, or merely a result of insufficient exposure to high-quality mathematical data. To explore this, we developed an automated method for generating high-quality, supervised mathematical datasets. The method carefully mutates existing math problems, ensuring both diversity and validity of the newly generated problems. This is achieved by a neuro-symbolic data generation framework combining the intuitive informalization strengths of LLMs, and the precise symbolic reasoning of math solvers along with projected Markov chain Monte Carlo sampling in the highly-irregular symbolic space. Empirical experiments demonstrate the high quality of data generated by the proposed method, and that the LLMs, specifically LLaMA-2 and Mistral, when realigned with the generated data, surpass their state-of-the-art counterparts.

With the rapid progress of Multimodal LLMs, evaluating their mathematical reasoning capabilities has become an increasingly important research direction. In particular, visual-textual mathematical reasoning serves as a key indicator of an MLLM's ability to comprehend and solve complex, multi-step quantitative problems. While existing benchmarks such as MathVista and MathVerse have advanced the evaluation of multimodal math proficiency, they primarily rely on digitally rendered content and fall short in capturing the complexity of real-world scenarios. To bridge this gap, we introduce MathScape, a novel benchmark focused on assessing MLLMs' reasoning ability in realistic mathematical contexts. MathScape comprises 1,369 high-quality math problems paired with human-captured real-world images, closely reflecting the challenges encountered in practical educational settings. We conduct a thorough multi-dimensional evaluation across nine leading closed-source MLLMs, three open-source MLLMs with over 20 billion parameters, and seven smaller-scale MLLMs. Our results show that even SOTA models struggle with real-world math tasks, lagging behind human performance – highlighting critical limitations in current model capabilities. Moreover, we find that strong performance on synthetic or digitally rendered images does not guarantee similar effectiveness on real-world tasks. This underscores the necessity of MathScape in the next stage of multimodal mathematical reasoning.

Large language models (LLMs) have demonstrated remarkable performance on a variety of natural language tasks based on just a few examples of natural language instructions, reducing the need for extensive feature engineering. However, most powerful LLMs are closed-source or limited in their capability for languages other than English. In this technical report, we present Baichuan 2, a series of large-scale multilingual language models containing 7 billion and 13 billion parameters, trained from scratch, on 2.6 trillion tokens. Baichuan 2 matches or outperforms other open-source models of similar size on public benchmarks like MMLU, CMMLU, GSM8K, and HumanEval. Furthermore, Baichuan 2 excels in vertical domains such as medicine and law. We will release all pre-training model checkpoints to benefit the research community in better understanding the training dynamics of Baichuan 2.