HM
Haitao Mi
cs.CLcs.AIcs.LGcs.CVcs.SEcs.CRcs.GTcs.HCcs.IRcs.LO
On Valency
published · living versionsW_btf7zyns·v1 · currentpublished
Harness Self-Evolution for Long-Horizon Terminal Agents
with Ruhan Wang, Yucheng Shi, Zongxia Li, Junyao Yang +4
1 version
W_8n8b89ac·v1published
Preprints & journals
96 papers in the corpus · 2016–2026Harness Self-Evolution for Long-Horizon Terminal Agents2610.27435v1 · Ruhan Wang, Yucheng Shi, Zongxia Li et al.2026 · 0 citationsarXivon Valency
Long-Horizon-Terminal-Bench 2.0: Automating Diverse Task Construction and Measuring Sustained Agent Progress2610.14146v1 · Zongxia Li, Zhongzhi Li, Yucheng Shi et al.2026 · 0 citationsarXiv
Self-Supervised Scaling of Terminal Environments for Scientific Domains2610.02710v1 · Zhongzhi Li, Yucheng Shi, Zongxia Li et al.2026 · 0 citationsarXiv
Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite2610.02826v1 · Zongxia Li, Yucheng Shi, Zhongzhi Li et al.2026 · 0 citationsarXiv
SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking2609.13141v1 · Zhiwei Li, Lei Zhu, Hao Gu et al.2026 · 0 citationsarXivon Valency
T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks2609.11042v1 · Junyao Yang, Yucheng Shi, Zhongzhi Li et al.2026 · 0 citationsarXiv
FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience2609.03241v1 · Zixun Huang, Kishan Panaganti, Haitao Mi et al.2026 · 0 citationsarXiv
Recursive Synthesis for Long-Horizon Terminal Tasks2608.05466v3 · Zhongzhi Li, Yucheng Shi, Zongxia Li et al.2026 · 0 citationsarXiv
Stale but Stable: Staleness-Adaptive Trust Regions for Stabilizing Asynchronous Reinforcement Learning2607.18722v3 · Junyao Yang, Yucheng Shi, Zongxia Li et al.2026 · 0 citationsarXiv
Harness Handbook: Making Evolving Agent Harnesses Readable,Navigable, and Editable2607.13285v1 · Ruhan Wang, Yucheng Shi, Zongxia Li et al.2026 · 0 citationsarXiv
DeepSearch-World: Self-Distillation for Deep Search Agents in a Verifiable Environment2607.07820v2 · Xinyu Geng, Xuanhua He, Sixiang Chen et al.2026 · 0 citationsarXiv
Hierarchical Sparse Attention Done Right: Toward Infinite Context Modeling2607.02980v1 · Xiang Hu, Xinyu Wei, Hao Gu et al.2026 · 0 citationsarXiv
FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention2606.09079v3 · Yan Wang, Qifan Zhang, Jiachen Yu et al.2026 · 0 citationsarXiv
Learning to Build the Environment: Self-Evolving Reasoning RL via Verifiable Environment Synthesis2605.14392v1 · Yucheng Shi, Zhenwen Liang, Kishan Panaganti et al.2026 · 0 citationsarXiv
Reinforcing Multimodal Reasoning Against Visual Degradation2605.09262v1 · Rui Liu, Dian Yu, Haolin Liu et al.2026 · 0 citationsarXiv
DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification2605.09269v1 · Rui Liu, Dian Yu, Zhenwen Liang et al.2026 · 0 citationsarXiv
Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding2604.21268v1 · Wenkai Wang, Xiyun Li, Hongcan Guo et al.2026 · 0 citationsarXiv
Training LLM Agents for Spontaneous, Reward-Free Self-Evolution via World Knowledge Exploration2604.18131v1 · Qifan Zhang, Dongyang Ma, Tianqing Fang et al.2026 · 0 citationsarXiv
Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data2604.18493v1 · Zhenwen Liang, Yujun Zhou, Sidi Lu et al.2026 · 0 citationsarXiv
The Pensieve Paradigm: Stateful Language Models Mastering Their Own Context2602.12108v1 · Xiaoyuan Liu, Tian Liang, Dongyang Ma et al.2026 · 0 citationsarXiv
Free(): Learning to Forget in Malloc-Only Reasoning Models2602.08030v2 · Yilun Zheng, Dongyang Ma, Tian Liang et al.2026 · 0 citationsarXiv
Locas: Your Models are Principled Initializers of Locally-Supported Parametric Memories2602.05085v1 · Sidi Lu, Zhenwen Liang, Dongyang Ma et al.2026 · 0 citationsarXiv
Verified Critical Step Optimization for LLM Agents2602.03412v2 · Mukai Li, Qingcheng Zeng, Tianqing Fang et al.2026 · 0 citationsarXiv
Exploring Information Seeking Agent Consolidation2602.00585v2 · Guochen Yan, Jialong Wu, Zhengwei Tao et al.2026 · 0 citationsarXiv
Group Distributionally Robust Optimization-Driven Reinforcement Learning for LLM Reasoning2601.19280v1 · Kishan Panaganti, Zhenwen Liang, Wenhao Yu et al.2026 · 0 citationsarXiv
Save the Good Prefix: Precise Error Penalization via Process-Supervised RL to Enhance LLM Reasoning2601.18984v1 · Haolin Liu, Dian Yu, Sidi Lu et al.2026 · 0 citationsarXiv
Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification2601.15808v2 · Yuxuan Wan, Tianqing Fang, Zaitang Li et al.2026 · 1 citationarXiv
RAGShaper: Eliciting Sophisticated Agentic RAG Skills via Automated Data Synthesis2601.08699v1 · Zhengwei Tao, Bo Li, Jialong Wu et al.2026 · 0 citationsarXiv
DocDancer: Towards Agentic Document-Grounded Information Seeking2601.05163v1 · Qintong Zhang, Xinjie Lv, Jialong Wu et al.2026 · 0 citationsarXiv
Stable and Efficient Single-Rollout RL for Multimodal Reasoning2512.18215v1 · Rui Liu, Dian Yu, Lei Ke et al.2025 · 0 citationsarXiv
Guided Self-Evolving LLMs with Minimal Human Supervision2512.02472v1 · Wenhao Yu, Zhenwen Liang, Chengsong Huang et al.2025 · 0 citationsarXiv
SciAgent: A Unified Multi-Agent System for Generalistic Scientific Reasoning2511.08151v2 · Xuchen Li, Ruitao Wu, Xuanbo Liu et al.2025 · 0 citationsarXiv
DeepCompress: A Dual Reward Strategy for Dynamically Exploring and Compressing Reasoning Chains2510.27419v2 · Tian Liang, Wenxiang Jiao, Zhiwei He et al.2025 · 0 citationsarXiv
Every Question Has Its Own Value: Reinforcement Learning with Explicit Human Values2510.20187v1 · Dian Yu, Yulai Zhao, Kishan Panaganti et al.2025 · 0 citationsarXiv
WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models2510.14438v2 · Rui Wang, Ce Zhang, Jun-Yu Ma et al.2025 · 0 citationsarXiv
CLUE: Non-parametric Verification from Experience via Hidden-State Clustering2510.01591v1 · Zhenwen Liang, Ruosen Li, Yujun Zhou et al.2025 · 0 citationsarXiv
Dual-Uncertainty Guided Policy Learning for Multimodal Reasoning2510.01444v3 · Rui Liu, Dian Yu, Tong Zheng et al.2025 · 0 citationsarXiv
UniGist: Towards General and Hardware-aligned Sequence-level Long Context Compression2509.15763v1 · Chenlong Deng, Zhisong Zhang, Kelong Mao et al.2025 · 0 citationsarXiv
Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation2509.15194v4 · Yujun Zhou, Zhenwen Liang, Haolin Liu et al.2025 · 0 citationsarXiv
EconProver: Towards More Economical Test-Time Scaling for Automated Theorem Proving2509.12603v1 · Mukai Li, Linfeng Song, Zhenwen Liang et al.2025 · 0 citationsarXiv
Cognitive Kernel-Pro: A Framework for Deep Research Agents and Agent Foundation Models Training2508.00414v3 · Tianqing Fang, Zhisong Zhang, Xiaoyang Wang et al.2025 · 0 citationsarXiv
One Token to Fool LLM-as-a-Judge2507.08794v3 · Yulai Zhao, Haolin Liu, Dian Yu et al.2025 · 1 citationarXiv
MobileGUI-RL: Advancing Mobile GUI Agent through Reinforcement Learning in Online Environment2507.05720v1 · Yucheng Shi, Wenhao Yu, Zaitang Li et al.2025 · 0 citationsarXiv
Towards Solving More Challenging IMO Problems via Decoupled Reasoning and Proving2507.06804v1 · Zhenwen Liang, Linfeng Song, Yang Li et al.2025 · 0 citationsarXiv
DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural Language and Reinforcement Learning2505.23754v2 · Ziyin Zhang, Jiahao Xu, Zhiwei He et al.2025 · 0 citationsarXiv
InComeS: Integrating Compression and Selection Mechanisms into LLMs for Efficient Model Editing2505.22156v4 · Shuaiyi Li, Zhisong Zhang, Yang Deng et al.2025 · 0 citationsarXiv
VScan: Rethinking Visual Token Reduction for Efficient Large Vision-Language Models2505.22654v3 · Ce Zhang, Kaixin Ma, Tianqing Fang et al.2025 · 0 citationsarXiv
WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback2505.20013v2 · Minda Hu, Tianqing Fang, Jianshu Zhang et al.2025 · 1 citationarXiv
Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training2505.14681v2 · Mengru Wang, Xingyu Chen, Yue Wang et al.2025 · 0 citationsarXiv
Trust, But Verify: A Self-Verification Approach to Reinforcement Learning with Verifiable Rewards2505.13445v1 · Xiaoyuan Liu, Tian Liang, Zhiwei He et al.2025 · 2 citationsarXiv
Career total: 150 works. 96 are in this corpus.Showing the 50 most recent.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.