I am a master’s student at FudanNLP Lab, Fudan University, co-advised by Associate Prof. Tao Gui (桂 韬) and Prof. Xuanjing Huang (黄萱菁). My research focuses on self-evolving agents and reinforcement learning. I have end-to-end experience across the agent post-training stack — SFT, RL, and Harness Engineering. I am currently a Research Intern at Shanghai AI Laboratory, where I am Tech Lead for ML coding on Atria Dawn Preview. I previously interned at NEX-AGI (Shanghai Qiji Zhifeng Co., Ltd.) and at the Computing Research Department of Huawei. I received my B.S. from Fudan University in 2025.我是复旦大学 FudanNLP 实验室的硕士研究生，师从桂韬副教授与黄萱菁教授。我的研究方向为自进化智能体（self-evolving agents）与强化学习。我在智能体后训练全链路上拥有端到端经验——SFT、RL 以及 Harness 工程。我目前在上海人工智能实验室担任研究实习生，是 Atria Dawn Preview 的 ML coding 方向 Tech Lead。此前曾在 NEX-AGI（上海奇绩智峰）与华为计算研究部门实习。我于 2025 年在复旦大学获得学士学位。

You can find my publications on Google Scholar. Feel free to reach out to me at 1187524561@qq.com.你可以在 Google Scholar 上查看我的论文。欢迎通过邮箱 1187524561@qq.com 与我联系。

🔥 News最新动态

• 2026.09: 🎉🎉 Atria Dawn Preview, our agentic foundation model for scientific research and engineering workflows, is released on arXiv!🎉🎉 Atria Dawn Preview——我们面向科研与工程工作流的智能体基础模型，已在 arXiv 发布！
• 2026.05: 🎉🎉 AHE (Agentic Harness Engineering) is released on arXiv!🎉🎉 AHE（Agentic Harness Engineering）已在 arXiv 发布！
• 2026.04: 🎉🎉 MM-Doc-R1 is accepted by ACL 2026 Findings!🎉🎉 MM-Doc-R1 被 ACL 2026 Findings 录用！
• 2026.04: 🎉🎉 OctoBench is accepted by ACL 2026 Main!🎉🎉 OctoBench 被 ACL 2026 主会 录用！
• 2025.12: 🎉🎉 Nex-N1, our agentic model trained via a unified ecosystem for large-scale environment construction, is released on arXiv!🎉🎉 Nex-N1——我们通过统一生态进行大规模环境构建所训练的智能体模型，已在 arXiv 发布！

💻 Internships实习经历

• 2026.7 - Present · Research Intern, Shanghai AI Laboratory, Shanghai2026.7 - 至今 · 研究实习生，上海人工智能实验室，上海
  Tech Lead for ML coding on Atria Dawn PreviewAtria Dawn Preview 的 ML coding 方向 Tech Lead
  • Own training-data production for four target benchmarks — MLE-Bench, MLS-Bench, PaperBench and PostTrainBench — driving the whole pipeline of task synthesis, rollout and trajectory filtering: 8,300+ training tasks built and packaged in the Harbor format, large-scale rollout inside a GPU sandbox, and trajectories filtered against SFT data-quality standards, lifting Atria Dawn Preview’s ML coding ability into the first tier.负责 MLE-Bench、MLS-Bench、PaperBench、PostTrainBench 四个目标 benchmark 的训练数据生产，主导任务合成、rollout 与轨迹筛选全链路：累计构造 8,300+ 训练任务并统一封装为 Harbor 格式，在 GPU 沙箱中完成大规模 rollout，通过 SFT 数据质检标准过滤轨迹，最终支撑 Atria Dawn Preview 的 ML coding 能力达到第一梯队水平。
  • Also built a benchmark out of real research-user data, measuring how well a model can pick up a user’s request inside a live workspace and carry it forward.另基于真实科研用户数据构造 benchmark，评测模型在真实工作区上继续推进用户需求的能力。
• 2025.3 - 2026.6 · Research Intern, NEX-AGI (Shanghai Qiji Zhifeng Co., Ltd.), Shanghai2025.3 - 2026.6 · 研究实习生，NEX-AGI（上海奇绩智峰），上海
  Self-evolving agents & reinforcement learning for coding and search agents面向代码与搜索智能体的自进化智能体与强化学习
• 2024.7 - 2024.9 · Research Intern, Huawei Computing Product Line, Computing Research Department2024.7 - 2024.9 · 研究实习生，华为 计算产品线 计算研究部
  Fine-tuning long-context LLMs on Ascend 910B在昇腾 910B 上微调长上下文大模型
• 
  2024.4 - 2024.7 · Research Intern, FudanNLP Lab, Fudan University2024.4 - 2024.7 · 研究实习生，FudanNLP 实验室，复旦大学
  LLM safety: training red-team models for attack and defense大模型安全：训练用于攻防的红队模型

📝 Publications论文发表

Arxiv 2026

Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses

Jiahang Lin*, Shichun Liu*, Chengjun Pan*, Lizhi Lin, Shihan Dou, Xuanjing Huang, Hang Yan, Zhenhua Han†, Tao Gui†

• AHE is an observability stack for the automatic optimization of coding-agent harnesses, with three pillars: component observability (NexAU), experience observability (Agent Debugger), and decision observability (evidence-driven Evolve Agent).AHE 是一套用于自动优化编码智能体 harness 的可观测性体系，包含三大支柱：组件可观测性（NexAU）、经验可观测性（Agent Debugger）以及决策可观测性（基于证据的 Evolve Agent）。
• Without changing the model, AHE pushes Terminal-bench 2 from 69.7% to 77.0% across iterations, with strong cross-task and cross-model generalization.在不改动模型的前提下，AHE 通过多轮迭代将 Terminal-bench 2 从 69.7% 提升至 77.0%，并展现出强的跨任务、跨模型泛化能力。
• | | |

Arxiv 2026

OctoBench: Benchmarking Scaffold-Aware Instruction Following in Repository-Grounded Agentic Coding

Deming Ding*, Shichun Liu*, Enhui Yang*, Jiahang Lin*, Ziying Chen, Shihan Dou, Honglin Guo, Weiyu Cheng, Pengyu Zhao, Chengjun Xiao, Qunhong Zeng, Qi Zhang, Xuanjing Huang, Qidi Xu, Tao Gui

• Modern coding scaffolds turn LLMs into capable software agents, but their ability to follow scaffold-specified instructions remains under-examined. OctoBench benchmarks scaffold-aware instruction following in repository-grounded agentic coding.现代编码脚手架（scaffold）能把大模型变成强大的软件智能体，但它们遵循脚手架指定指令的能力仍缺乏系统评估。OctoBench 在仓库级智能体编码场景下，评测面向脚手架的指令遵循能力。
• OctoBench includes 34 environments and 217 tasks under three scaffold types, with 7,098 objective checklist items. We release an automated observation-and-scoring toolkit for full trajectories and fine-grained checks.OctoBench 包含三类脚手架下的 34 个环境、217 个任务，以及 7,098 条客观检查项。我们开源了一套可对完整轨迹进行自动观测与打分、支持细粒度检查的工具包。
• | | |

Arxiv 2025

Nex-N1: Agentic Models Trained via a Unified Ecosystem for Large-Scale Environment Construction

Nex-AGI Team: Yuxuan Cai, Lu Chen, …, Jiahang Lin, …, Xuanjing Huang, Xipeng Qiu

• We introduce a comprehensive method designed to systematically scale the diversity and complexity of interactive environments through three orthogonal dimensions: Complexity (NexAU), Diversity (NexA4A), and Fidelity (NexGAP).我们提出一套系统性方法，从三个正交维度规模化提升交互环境的多样性与复杂度：复杂度（NexAU）、多样性（NexA4A）与保真度（NexGAP）。
• Nex-N1 consistently outperforms SOTA open-source models and achieves competitive performance against frontier proprietary models on complex agentic tasks (SWE-bench, tau2).在复杂智能体任务（SWE-bench、tau2）上，Nex-N1 持续超越 SOTA 开源模型，并在与前沿闭源模型的对比中展现出有竞争力的表现。
• |

Arxiv 2026

MM-Doc-R1: Training Agents for Long Document Visual Question Answering through Multi-turn Reinforcement Learning

Jiahang Lin*, Kai Hu*, Binghai Wang, Yuhao Zhou, Zhiheng Xi, Honglin Guo, Shichun Liu, Junzhe Wang, Shihan Dou, Enyu Zhou, Hang Yan, Zhenhua Han, Tao Gui†, Qi Zhang†, Xuanjing Huang†

• Conventional RAG systems struggle with complex multi-hop queries over long documents due to their single-pass retrieval. MM-Doc-R1 trains agents for long-document visual question answering via multi-turn reinforcement learning.传统 RAG 系统受限于单次检索，难以应对长文档上的复杂多跳查询。MM-Doc-R1 通过多轮强化学习训练面向长文档视觉问答的智能体。
• The agent learns to interleave retrieval and reasoning across many turns, achieving substantial gains on long-document VQA benchmarks compared to single-pass and prompt-only baselines.该智能体学会在多轮之间交替进行检索与推理，相比单次检索与仅提示（prompt-only）基线，在长文档 VQA 基准上取得显著提升。
• 

Arxiv 2026

EVPO: Explained Variance Policy Optimization for Adaptive Critic Utilization in LLM Post-Training

Chengjun Pan*, Shichun Liu*, Jiahang Lin*, Dingwei Zhu, Jiazheng Zhang, Shihan Dou, Songyang Gao, Zhenhua Han, Binghai Wang, Rui Zheng, Xuanjing Huang†, Tao Gui†, Yansong Feng†

• We cast baseline selection in LLM post-training as a Kalman filtering problem, unifying PPO and GRPO as two extremes of the Kalman gain, and prove that the sign of explained variance (EV) is the exact boundary separating the variance-reducing from the variance-inflating critic regime.我们将大模型后训练中的 baseline 选择建模为卡尔曼滤波问题，把 PPO 与 GRPO 统一为卡尔曼增益的两个极端，并证明 explained variance（EV）的符号正是区分「降方差」与「增方差」critic 区间的精确边界。
• EVPO adaptively switches between critic-based and batch-mean advantage estimation per step based on EV sign, achieving the best results across Sokoban, FrozenLake, WebShop, and MATH.EVPO 依据每一步的 EV 符号，在基于 critic 与基于 batch 均值的优势估计之间自适应切换，在 Sokoban、FrozenLake、WebShop 与 MATH 上均取得最佳结果。
• |

Arxiv 2026

EvoIn: Bridging Evolution and Internalization for Agent Fine-Tuning

Shihan Dou*, Shaofan Liu*, Zhonghang Lu*, Jiahang Lin, Shichun Liu, Binghai Wang, Jiajie Jin, Guanting Dong, Tao Gui†, Qi Zhang, Xuanjing Huang

• EvoIn is an agent fine-tuning framework that bridges evolution and internalization: it analyzes execution traces to evolve and validate new decision-making procedures by temporarily instantiating them in the harness, then rewrites the resulting traces into self-contained reasoning so the procedures persist without the evolved harness at inference time.EvoIn 是一个打通「进化」与「内化」的智能体微调框架：先分析执行轨迹，通过把新的决策流程临时注入 harness 来进化并验证它们，再把由此得到的轨迹改写为自包含的推理，使这些流程在推理时脱离进化后的 harness 依然保留。
• Fine-tuning on the rewritten traces improves performance by 10.9 points in-domain and 9.2 points out-of-domain, and the internalized procedures generalize to unseen tasks — agents learn, for example, to adapt their verification effort to task complexity.在改写后的轨迹上微调带来域内 10.9 分、域外 9.2 分的提升，且内化后的决策流程可泛化到未见任务——例如智能体会学到按任务复杂度调整验证力度。
• 

Arxiv 2026

Atria Dawn: The Dawn of Agentic Superintelligence

Honglin Guo, Tao Gui†, …, Jiahang Lin (core contributor), …, Xuanjing Huang, Yu-Gang Jiang, Bowen Zhou†

• Atria Dawn Preview is a foundation agentic language model for scientific research and engineering workflows, trained via a Verifiable Experience Pipeline that connects tool-mediated interactions to executable environments and externally verified outcomes.Atria Dawn Preview 是一个面向科研与工程工作流的智能体基础模型，通过 Verifiable Experience Pipeline 训练——该流水线把工具交互接入可执行环境，并以外部可验证的结果作为监督信号。
• Across 16 benchmarks spanning real-world research, engineering, and digital work, it is competitive with frontier agents and achieves the highest reported score on five of them. The paper also studies the human–AI collaboration behind the model itself, from 769 task records across 56 participants.在覆盖真实科研、工程与数字工作的 16 个基准上，它与前沿智能体相当，并在其中 5 个上取得已报告的最高分。论文还以该模型自身的研发过程为案例，基于 56 位参与者的 769 条任务记录分析了人机协作模式。
• 

🎖 Honors and Awards荣誉奖项

• 2024, 1st Prize, 2024 Tencent AI Arena Global Open Competition · Agent Game Algorithm Track · Mainland China Regional Final. Team: 五角场三分王, Fudan University.2024，一等奖，2024 腾讯 AI Arena 全球公开赛 · 智能体博弈算法赛道 · 中国大陆赛区总决赛。队伍：五角场三分王，复旦大学。
• 2022-2023, Third Prize, Fudan University Outstanding Student Scholarship.2022-2023 学年，复旦大学优秀学生奖学金三等奖。
• 2021-2022, “Star of Yangfan” of Fujian Province; Youth Talent Recruitment Volunteer of Fujian Province.2021-2022 学年，福建省扬帆之星；福建省青年引才志愿者。

📖 Education教育经历

• 2025.09 - Present, M.S., Fudan University (FudanNLP Lab).2025.09 - 至今，硕士，复旦大学（FudanNLP 实验室）。
• 2021.09 - 2025.06, B.S., Fudan University.2021.09 - 2025.06，学士，复旦大学。