21 citations · 47 across the 31 of their papers we have counts for
35 papers
Wrong Votes, Right Gradients: Why Majority-Vote Rewards Track Verifiable RL
Dylan P. Wilson, Chengsong Huang, Jiaxin Huang
Reinforcement learning with self-generated labels, such as majority vote over a model's own samples, often recovers most of the gain of training with verifiable rewards. This is pu…
Long-Horizon-Terminal-Bench 2.0: Automating Diverse Task Construction and Measuring Sustained Agent Progress
Zongxia Li, Zhongzhi Li, Yucheng Shi +14
Long-horizon terminal benchmarks measure whether an agent can sustain progress through extended workflows with process rewards, not only whether its final state passes a binary che…
Questioning the Questions: Sustaining Self-Evolution in Reasoning Models
Jinyuan Li, Chengsong Huang, Langlin Huang +4
Self-evolving reasoning models learn from their own generated questions, yet repeated self-training can lead to performance collapse. In this paper, we investigate why performance…
Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite
Zongxia Li, Yucheng Shi, Zhongzhi Li +7
Successful trajectories on difficult tasks provide valuable supervision for model improvement, but specialized harnesses introduce interventions that may be unavailable during depl…
RRSI: Regularized Recursive Self-Improvement of Agent Harnesses
Peng Xia, Rujun Han, Zifeng Wang +11
An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Rece…
EnvHarness: Awakening Static Worlds for Agent Learning
Chengsong Huang, Zifeng Wang, Rujun Han +14
LLM agents learn by interacting with environments, yet these environments are hand-built and static: blind to an agent's weaknesses, and quickly left behind as it improves. While r…
Small RL Controller, Large Language Model: RL-Guided Adaptive Sampling for Test-Time Scaling
Runpeng Dai, Tong Zheng, Rui Liu +2
Test-time scaling improves the reasoning performance of large language models but incurs substantial cost in both total computation and latency. Existing adaptive sampling methods…
Process Rewards with Learned Reliability
Jinyuan Li, Langlin Huang, Chengsong Huang +5
Process Reward Models (PRMs) provide step-level feedback for reasoning, but current PRMs usually output only a single reward score for each step. Downstream methods must therefore…
LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling
Tong Zheng, Haolin Liu, Chengsong Huang +10
Test-time scaling (TTS) has become an effective approach for improving large language model performance by allocating additional computation during inference. However, existing TTS…
G-Zero: Self-Play for Open-Ended Generation from Zero Data
Chengsong Huang, Haolin Liu, Tong Zheng +7
Self-evolving LLMs excel in verifiable domains but struggle in open-ended tasks, where reliance on proxy LLM judges introduces capability bottlenecks and reward hacking. To overcom…
Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration
Langlin Huang, Chengsong Huang, Jinyuan Li +3
Reinforcement learning with verifiable rewards, particularly Group Relative Policy Optimization (GRPO), has significantly advanced the reasoning capabilities of Large Language Mode…
MM-Zero: Self-Evolving Multi-Model Vision Language Models From Zero Data
Zongxia Li, Hongyang Du, Chengsong Huang +8
Self-evolving has emerged as a key paradigm for improving foundational models such as Large Language Models (LLMs) and Vision Language Models (VLMs) with minimal human intervention…
Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probing
Tong Zheng, Chengsong Huang, Runpeng Dai +9
Parallel thinking has emerged as a promising paradigm for reasoning, yet it imposes significant computational burdens. Existing efficiency methods primarily rely on local, per-traj…
Training Data Efficiency in Multimodal Process Reward Models
Jinyuan Li, Chengsong Huang, Langlin Huang +4
Multimodal Process Reward Models (MPRMs) are central to step-level supervision for visual reasoning in MLLMs. Training MPRMs typically requires large-scale Monte Carlo (MC)-annotat…
TTCS: Test-Time Curriculum Synthesis for Self-Evolving
Chengyi Yang, Zhishang Xiang, Yunbo Tang +5
Test-Time Training offers a promising way to improve the reasoning ability of large language models (LLMs) by adapting the model using only the test questions. However, existing me…
RelayLLM: Efficient Reasoning via Collaborative Decoding
Chengsong Huang, Tong Zheng, Langlin Huang +3
Large Language Models (LLMs) for complex reasoning is often hindered by high computational costs and latency, while resource-efficient Small Language Models (SLMs) typically lack t…
Benchmark^2: Systematic Evaluation of LLM Benchmarks
Qi Qian, Chengsong Huang, Jingwen Xu +13
The rapid proliferation of benchmarks for evaluating large language models (LLMs) has created an urgent need for systematic methods to assess benchmark quality itself. We propose B…
MoCo: A One-Stop Shop for Model Collaboration Research
Shangbin Feng, Yuyang Bai, Ziyuan Yang +17
Advancing beyond single monolithic language models (LMs), recent research increasingly recognizes the importance of model collaboration, where multiple LMs collaborate, compose, an…
Guided Self-Evolving LLMs with Minimal Human Supervision
Wenhao Yu, Zhenwen Liang, Chengsong Huang +4
AI self-evolution has long been envisioned as a path toward superintelligence, where models autonomously acquire, refine, and internalize knowledge from their own learning experien…
UniRel: Relation-Centric Knowledge Graph Question Answering with RL-Tuned LLM Reasoning
Yinxu Tang, Chengsong Huang, Jiaxin Huang +1
Knowledge Graph Question Answering (KGQA) has largely focused on entity-centric queries that return a single answer entity. However, many real-world questions are inherently relati…
VisPlay: Self-Evolving Vision-Language Models from Images
Yicheng He, Chengsong Huang, Zongxia Li +2
Reinforcement learning (RL) provides a principled framework for improving Vision-Language Models (VLMs) on complex reasoning tasks. However, existing RL approaches often rely on hu…
Parallel-R1: Towards Parallel Thinking via Reinforcement Learning
Tong Zheng, Hongming Zhang, Wenhao Yu +7
Parallel thinking has emerged as a novel approach for enhancing the reasoning capabilities of large language models (LLMs) by exploring multiple reasoning paths concurrently. Howev…
Self-Rewarding Vision-Language Model via Reasoning Decomposition
Zongxia Li, Wenhao Yu, Chengsong Huang +8
Vision-Language Models (VLMs) often suffer from visual hallucinations: generating things that are not consistent with visual inputs and language shortcuts, where they skip the visu…
R-Zero: Self-Evolving Reasoning LLM from Zero Data
Chengsong Huang, Wenhao Yu, Xiaoyang Wang +6
Self-evolving Large Language Models (LLMs) offer a scalable path toward super-intelligence by autonomously generating, refining, and learning from their own experiences. However, e…