Publications (64)
Teaching Thinking Models to Reason with Tools: A Full-Pipeline Recipe for Tool-Integrated Reasoning
Qianjia Cheng, Yuchen Zhang, Zhilin Wang +9
Tool-integrated reasoning (TIR) offers a direct way to extend thinking models beyond the limits of text-only reasoning. Paradoxically, we observe that tool-enabled evaluation can d…
UltraFeedback: Boosting Language Models with Scaled AI Feedback
Ganqu Cui, Lifan Yuan, Ning Ding +9
Learning from human feedback has become a pivot technique in aligning large language models (LLMs) with human preferences. However, acquiring vast and premium human feedback is bot…
PhysicsMinions: Winning Gold Medals in the Latest Physics Olympiads with a Coevolutionary Multimodal Multi-Agent System
Fangchen Yu, Junchi Yao, Ziyi Wang +10
Physics is central to understanding and shaping the real world, and the ability to solve physics problems is a key indicator of real-world physical intelligence. Physics Olympiads,…
A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond
Xiaoye Qu, Yafu Li, Zhao-Chen Su +15
Recent Large Reasoning Models (LRMs), such as DeepSeek-R1 and OpenAI o1, have demonstrated strong performance gains by scaling up the length of Chain-of-Thought (CoT) reasoning dur…
Noise Contrastive Alignment of Language Models with Explicit Rewards
Huayu Chen, Guande He, Lifan Yuan +3
User intentions are typically formalized as evaluation rewards to be maximized when fine-tuning language models (LMs). Existing alignment methods, such as Direct Preference Optimiz…
SCI-Verifier: Scientific Verifier with Thinking
Shenghe Zheng, Chenyu Huang, Fangchen Yu +8
As large language models (LLMs) are increasingly applied to scientific reasoning, the complexity of answer formats and the diversity of equivalent expressions make answer verificat…
V-GameGym: Visual Game Generation for Code Large Language Models
Wei Zhang, Jack Yang, Renshuai Tao +9
Code large language models have demonstrated remarkable capabilities in programming tasks, yet current benchmarks primarily focus on single modality rather than visual game develop…
Process Reinforcement through Implicit Rewards
Ganqu Cui, Lifan Yuan, Zefan Wang +22
Dense process rewards have proven a more effective alternative to the sparse outcome-level rewards in the inference-time scaling of large language models (LLMs), particularly in ta…
Exploring the Universal Vulnerability of Prompt-based Learning Paradigm
Lei Xu, Yangyi Chen, Ganqu Cui +2
Prompt-based learning paradigm bridges the gap between pre-training and fine-tuning, and works effectively under the few-shot setting. However, we find that this learning paradigm…
Post-Trained MoE Can Skip Half Experts via Self-Distillation
Xingtai Lv, Li Sheng, Kaiyan Zhang +12
Mixture-of-Experts (MoE) scales language models efficiently through sparse expert activation, and its dynamic variant further reduces computation by adjusting the activated experts…
MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
Tianyu Yu, Zefan Wang, Chongyi Wang +31
Multimodal Large Language Models (MLLMs) are undergoing rapid progress and represent the frontier of AI development. However, their training and inference efficiency have emerged a…
The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models
Ganqu Cui, Yuchen Zhang, Jiacheng Chen +14
This paper aims to overcome a major obstacle in scaling RL for reasoning with LLMs, namely the collapse of policy entropy. Such phenomenon is consistently observed across vast RL r…
Scalable Efficient Training of Large Language Models with Low-dimensional Projected Attention
Xingtai Lv, Ning Ding, Kaiyan Zhang +3
Improving the effectiveness and efficiency of large language models (LLMs) simultaneously is a critical yet challenging research goal. In this paper, we find that low-rank pre-trai…
Advancing LLM Reasoning Generalists with Preference Trees
Lifan Yuan, Ganqu Cui, Hanbin Wang +12
We introduce Eurus, a suite of large language models (LLMs) optimized for reasoning. Finetuned from Mistral-7B and CodeLlama-70B, Eurus models achieve state-of-the-art results amon…
From Adversarial Arms Race to Model-centric Evaluation: Motivating a Unified Automatic Robustness Evaluation Framework
Yangyi Chen, Hongcheng Gao, Ganqu Cui +10
Textual adversarial attacks can discover models' weaknesses by adding semantic-preserved but misleading perturbations to the inputs. The long-lasting adversarial attack-and-defense…
The Right Time Matters: Data Arrangement Affects Zero-Shot Generalization in Instruction Tuning
Bingxiang He, Ning Ding, Cheng Qian +10
Understanding alignment techniques begins with comprehending zero-shot generalization brought by instruction tuning, but little of the mechanism has been understood. Existing work…
Why Should Adversarial Perturbations be Imperceptible? Rethink the Research Paradigm in Adversarial NLP
Yangyi Chen, Hongcheng Gao, Ganqu Cui +4
Textual adversarial samples play important roles in multiple subfields of NLP research, including security, evaluation, explainability, and data augmentation. However, most work mi…
MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
Shengding Hu, Yuge Tu, Xu Han +22
The burgeoning interest in developing Large Language Models (LLMs) with up to trillion parameters has been met with concerns regarding resource efficiency and practical expense, pa…
Teaching Large Reasoning Models Effective Reflection
Hanbin Wang, Jingwei Song, Jinpeng Li +5
Large Reasoning Models (LRMs) have recently shown impressive performance on complex reasoning tasks, often by engaging in self-reflective behaviors such as self-critique and backtr…
Achieving Gold-Medal-Level Olympiad Reasoning via Simple and Unified Scaling
Yafu Li, Runzhe Zhan, Haoran Zhang +25
Recent progress in reasoning models has substantially advanced long-horizon mathematical and scientific problem solving, with several systems now reaching gold-medal-level performa…
Intern-S1-Pro: Scientific Multimodal Foundation Model at Trillion Scale
Yicheng Zou, Dongsheng Zhu, Lin Zhu +174
We introduce Intern-S1-Pro, the first one-trillion-parameter scientific multimodal foundation model. Scaling to this unprecedented size, the model delivers a comprehensive enhancem…
Learning to Reason under Off-Policy Guidance
Jianhao Yan, Yafu Li, Zican Hu +5
Recent advances in large reasoning models (LRMs) demonstrate that sophisticated behaviors such as multi-step reasoning and self-reflection can emerge via reinforcement learning wit…
Draft-OPD: On-Policy Distillation for Speculative Draft Models
Haodi Lei, Yafu Li, Haoran Zhang +8
Speculative decoding accelerates large language model inference by pairing a target model with a lightweight draft model whose proposed tokens are verified in parallel. A common wa…
TEMPO: Scaling Test-time Training for Large Reasoning Models
Qingyang Zhang, Xinke Kong, Haitao Wu +7
Test-time training (TTT) adapts model parameters on unlabeled test instances during inference time, which continuously extends capabilities beyond the reach of offline training. De…
Evaluating Modules in Graph Contrastive Learning
Ganqu Cui, Yufeng Du, Cheng Yang +5
The recent emergence of contrastive learning approaches facilitates the application on graph representation learning (GRL), introducing graph contrastive learning (GCL) into the li…
JustRL: Scaling a 1.5B LLM with a Simple RL Recipe
Bingxiang He, Zekai Qu, Zeyuan Liu +9
Recent advances in reinforcement learning for large language models have converged on increasing complexity: multi-stage training pipelines, dynamic hyperparameter schedules, and c…
SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning
Haozhan Li, Yuxin Zuo, Jiale Yu +18
Vision-Language-Action (VLA) models have recently emerged as a powerful paradigm for robotic manipulation. Despite substantial progress enabled by large-scale pretraining and super…
Think Longer to Explore Deeper: Learn to Explore In-Context via Length-Incentivized Reinforcement Learning
Futing Wang, Jianhao Yan, Yun Luo +6
Achieving effective test-time scaling requires models to engage in In-Context Exploration -- the intrinsic ability to generate, verify, and refine multiple reasoning hypotheses wit…
Wisdom of the Crowd: Reinforcement Learning from Coevolutionary Collective Feedback
Wenzhen Yuan, Shengji Tang, Weihao Lin +8
Reinforcement learning (RL) has significantly enhanced the reasoning capabilities of large language models (LLMs), but its reliance on expensive human-labeled data or complex rewar…
Tool Learning with Foundation Models
Yujia Qin, Shengding Hu, Yankai Lin +38
Humans possess an extraordinary ability to create and utilize tools, allowing them to overcome physical limitations and explore new frontiers. With the advent of foundation models,…
Prototypical Verbalizer for Prompt-based Few-shot Tuning
Ganqu Cui, Shengding Hu, Ning Ding +2
Prompt-based tuning for pre-trained language models (PLMs) has shown its effectiveness in few-shot learning. Typically, prompt-based tuning wraps the input text into a cloze questi…
FlowRL: Matching Reward Distributions for LLM Reasoning
Xuekai Zhu, Daixuan Cheng, Dinghuai Zhang +20
We propose FlowRL: matching the full reward distribution via flow balancing instead of maximizing rewards in large language model (LLM) reinforcement learning (RL). Recent advanced…
INTERVENOR: Prompting the Coding Ability of Large Language Models with the Interactive Chain of Repair
Hanbin Wang, Zhenghao Liu, Shuo Wang +4
This paper introduces INTERVENOR (INTERactiVE chaiN Of Repair), a system designed to emulate the interactive code repair processes observed in humans, encompassing both code diagno…
A Survey of Reinforcement Learning for Large Reasoning Models
Kaiyan Zhang, Yuxin Zuo, Bingxiang He +36
In this paper, we survey recent advances in Reinforcement Learning (RL) for reasoning with Large Language Models (LLMs). RL has achieved remarkable success in advancing the frontie…
RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback
Tianyu Yu, Yuan Yao, Haoye Zhang +8
Multimodal Large Language Models (MLLMs) have recently demonstrated impressive capabilities in multimodal understanding, reasoning, and interaction. However, existing MLLMs prevale…
P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics Olympiads
Yun Luo, Futing Wang, Qianjia Cheng +28
The transition from symbolic manipulation to science-grade reasoning represents a pivotal frontier for Large Language Models (LLMs), with physics serving as the critical test ancho…
From and to : LLMs Learn New Skills in RL by Composing Old Ones
Lifan Yuan, Weize Chen, Yuchen Zhang +7
Does RL teach LLMs genuinely new skills, or does it merely activate existing ones? This question lies at the core of ongoing debates about the role of RL in LLM post-training. On o…
A Close Look into the Calibration of Pre-trained Language Models
Yangyi Chen, Lifan Yuan, Ganqu Cui +2
Pre-trained language models (PLMs) may fail in giving reliable estimates of their predictive uncertainty. We take a close look into this problem, aiming to answer two questions: (1…
A Unified Evaluation of Textual Backdoor Learning: Frameworks and Benchmarks
Ganqu Cui, Lifan Yuan, Bingxiang He +3
Textual backdoor attacks are a kind of practical threat to NLP systems. By injecting a backdoor in the training phase, the adversary could control model predictions via predefined…
Controllable Preference Optimization: Toward Controllable Multi-Objective Alignment
Yiju Guo, Ganqu Cui, Lifan Yuan +9
Alignment in artificial intelligence pursues the consistency between model responses and human preferences as well as values. In practice, the multifaceted nature of human preferen…
P1: Mastering Physics Olympiads with Reinforcement Learning
Jiacheng Chen, Qianjia Cheng, Fangchen Yu +25
Recent progress in large language models (LLMs) has moved the frontier from puzzle-solving to science-grade reasoning-the kind needed to tackle problems whose answers must stand ag…
NFT: Bridging Supervised Learning and Reinforcement Learning in Math Reasoning
Huayu Chen, Kaiwen Zheng, Qinsheng Zhang +8
Reinforcement Learning (RL) has played a central role in the recent surge of LLMs' math abilities by enabling self-improvement through binary verifier signals. In contrast, Supervi…
InCoder-32B: Code Foundation Model for Industrial Scenarios
Jian Yang, Wei Zhang, Jiajun Wu +25
Recent code large language models have achieved remarkable progress on general programming tasks. Nevertheless, their performance degrades significantly in industrial scenarios tha…
RLPR: Extrapolating RLVR to General Domains without Verifiers
Tianyu Yu, Bo Ji, Shouli Wang +9
Reinforcement Learning with Verifiable Rewards (RLVR) demonstrates promising potential in advancing the reasoning capabilities of LLMs. However, its success remains largely confine…
RLAIF-V: Open-Source AI Feedback Leads to Super GPT-4V Trustworthiness
Tianyu Yu, Haoye Zhang, Qiming Li +13
Traditional feedback learning for hallucination reduction relies on labor-intensive manual labeling or expensive proprietary models. This leaves the community without foundational…
TTRL: Test-Time Reinforcement Learning
Yuxin Zuo, Kaiyan Zhang, Li Sheng +13
This paper investigates Reinforcement Learning (RL) on data without explicit labels for reasoning tasks in Large Language Models (LLMs). The core challenge of the problem is reward…
Graph Neural Networks: A Review of Methods and Applications
Jie Zhou, Ganqu Cui, Shengding Hu +6
Lots of learning tasks require dealing with graph data which contains rich relation information among elements. Modeling physics systems, learning molecular fingerprints, predictin…
HiPhO: How Far Are (M)LLMs from Humans in the Latest High School Physics Olympiad Benchmark?
Fangchen Yu, Haiyuan Wan, Qianjia Cheng +14
Recently, the physical capabilities of (M)LLMs have garnered increasing attention. However, existing benchmarks for physics suffer from two major gaps: they neither provide systema…
How Far Can Unsupervised RLVR Scale LLM Training?
Bingxiang He, Yuxin Zuo, Zeyuan Liu +18
Unsupervised reinforcement learning with verifiable rewards (URLVR) offers a pathway to scale LLM training beyond the supervision bottleneck by deriving rewards without ground trut…
A Roadmap for Big Model
Sha Yuan, Hanyu Zhao, Shuai Zhao +97
With the rapid development of deep learning, training Big Models (BMs) for multiple downstream tasks becomes a popular paradigm. Researchers have achieved various outcomes in the c…
Few-shot Classification with Hypersphere Modeling of Prototypes
Ning Ding, Yulin Chen, Ganqu Cui +4
Metric-based meta-learning is one of the de facto standards in few-shot learning. It composes of representation learning and metrics calculation designs. Previous works construct c…
Free Process Rewards without Process Labels
Lifan Yuan, Wendi Li, Huayu Chen +6
Different from its counterpart outcome reward models (ORMs), which evaluate the entire responses, a process reward model (PRM) scores a reasoning trajectory step by step, providing…
UltraIF: Advancing Instruction Following from the Wild
Kaikai An, Li Sheng, Ganqu Cui +4
Instruction-following made modern large language models (LLMs) helpful assistants. However, the key to taming LLMs on complex instructions remains mysterious, for that there are hu…
Scaling Physical Reasoning with the PHYSICS Dataset
Shenghe Zheng, Qianjia Cheng, Junchi Yao +9
Large Language Models (LLMs) have achieved remarkable progress on advanced reasoning tasks such as mathematics and coding competitions. Meanwhile, physics, despite being both reaso…
AIR: A Systematic Analysis of Annotations, Instructions, and Response Pairs in Preference Dataset
Bingxiang He, Wenbin Zhang, Jiaxi Song +11
Preference learning is critical for aligning large language models (LLMs) with human values, yet its success hinges on high-quality datasets comprising three core components: Prefe…
UltraMedical: Building Specialized Generalists in Biomedicine
Kaiyan Zhang, Sihang Zeng, Ermo Hua +11
Large Language Models (LLMs) have demonstrated remarkable capabilities across various domains and are moving towards more specialized areas. Recent advanced proprietary models such…
Intern-S1: A Scientific Multimodal Foundation Model
Lei Bai, Zhongrui Cai, Yuhang Cao +173
In recent years, a plethora of open-source foundation models have emerged, achieving remarkable progress in some widely attended fields, with performance being quite close to that…
InternAgentHarness: A Scalable Synthetic Environment for Enhancing LLM Agentic Abilities
Peiji Li, Jiasheng Ye, Yongkang Chen +19
Large language models (LLMs) are increasingly expected to act as generalist agents capable of solving complex real-world problems. Training such agents, however, requires stable an…
Decoder Tuning: Efficient Language Understanding as Decoding
Ganqu Cui, Wentao Li, Ning Ding +3
With the evergrowing sizes of pre-trained models (PTMs), it has been an emerging practice to only provide the inference APIs for users, namely model-as-a-service (MaaS) setting. To…
Adaptive Graph Encoder for Attributed Graph Embedding
Ganqu Cui, Jie Zhou, Cheng Yang +1
Attributed graph embedding, which learns vector representations from graph topology and node features, is a challenging task for graph analysis. Recently, methods based on graph co…
MiniCPM4: Ultra-Efficient LLMs on End Devices
MiniCPM Team, Chaojun Xiao, Yuxuan Li +80
This paper introduces MiniCPM4, a highly efficient large language model (LLM) designed explicitly for end-side devices. We achieve this efficiency through systematic innovation in…
From Drafts to Answers: Unlocking LLM Potential via Aggregation Fine-Tuning
Yafu Li, Zhilin Wang, Tingchen Fu +3
Scaling data and model size has been proven effective for boosting the performance of large language models. In addition to training-time scaling, recent studies have revealed that…
Mastering Text, Code and Math Simultaneously via Fusing Highly Specialized Language Models
Ning Ding, Yulin Chen, Ganqu Cui +6
Underlying data distributions of natural language, programming code, and mathematical symbols vary vastly, presenting a complex challenge for large language models (LLMs) that stri…
Revisiting Out-of-distribution Robustness in NLP: Benchmark, Analysis, and LLMs Evaluations
Lifan Yuan, Yangyi Chen, Ganqu Cui +6
This paper reexamines the research on out-of-distribution (OOD) robustness in the field of NLP. We find that the distribution shift settings in previous studies commonly lack adequ…