I’m a second year master student from Tsinghua University (THU), Shenzhen Graduate School. My advisor is Prof. Yujiu Yang. Prior to this, I obtained my bachelor’s degree from the Department of Electronic Engineering at Tsinghua University in June 2025.

I am currently interning in Tencent AI Lab, now Hunyuan Frontier, focus on memory compression in long context scenarios. Before this, I interned in ByteDance, focusing on post-training of MLLM for e-commerce scenarios. Prior to this, I interned in TeleAI, working on the post-training of the Telechat series foundation models. I have also interned in Tencent before for kernel development.

My research interest includes Reward Model/LLM-as-a-Judge and LLM/MLLM post training.

Activities

• 2026.9 🎉 Think-with-rubrics: From external evaluator to internal reasoning guidance is accepted by NeurIPS 2026 Main Track!
• 2026.8 🎉 Two papers are accepted by EMNLP 2026 Findings!
• 2026.6 🎉 Our new work FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention has been posted on arxiv!
• 2025.12 🧐 I start my internship in Tencent AI Lab, focusing on long-context compression.
• 2025.9 📖 I enroll in the Shenzhen Graduate School of Tsinghua University to pursue a Master degree, with an expected graduation date in 2028.
• 2025.8 🎉 Improve llm-as-a-judge ability as a general ability is accepted by EMNLP 2025 Main Conference. Welcome to use our SOTA LLM-as-a-Judge model RISE-Judge!
• 2025.6 🎆 I graduate from Department of Electronic Engineering, Tsinghua University and get bachelor degree.
• 2025.4 💼 I begin an internship at ByteDance, focusing on MLLM post-training research.

Publications

• Think-with-rubrics: From external evaluator to internal reasoning guidance
  • Jiachen Yu^*, Zhihao Xu^*, Junjie Wang, Yujiu Yang
  • NeurIPS 2026 Main Track
• S2J: Bridging the Gap Between Solving and Judging Ability in Generative Reward Models
  • Shaoning Sun^*, Jiachen Yu^*, Zongqi Wang, Xuewei Yang, Tianle Gu, Yujiu Yang
  • EMNLP 2026 Findings
• Internalize the Temperature: On-Policy Self-Distillation as Policy Reheater for Reinforcement Learning
  • Xuewei Yang^*, Jiachen Yu^*, Jie Wu, Shaoning Sun, Junjie Wang, Yujiu Yang
  • EMNLP 2026 Findings
• Improve llm-as-a-judge ability as a general ability
  • Jiachen Yu^*, Shaoning Sun^*, Xiaohui Hu, Jiaxu Yan, Kaidong Yu, Xuelong Li
  • EMNLP 2025 Main Conference
  • Finished during internship in TeleAI.

Technical Reports

• FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention
  • Yan Wang^*, Qifan Zhang^*, Jiachen Yu^*, Tian Liang^*, Dongyang Ma^*, Xiang Hu, Zibo Lin, Chunyang Li, Zhichao Wang, Miao Peng, Nuo Chen, Jia Li, Yujiu Yang, Haitao Mi, Dong Yu
  • Finished during internship in Tencent Hunyuan Frontier

Preprints

• CoWork-X: Experience-Optimized Co-Evolution for Multi-Agent Collaboration System
  • Zexin Lin^*, Jiachen Yu^*, Haoyang Zhang, Yuzhao Li, Zhonghang Li, Yujiu Yang, Junjie Wang, Xiaoqiang Ji
• VFaith: Do Large Multimodal Models Really Reason on Seen Images Rather than Previous Memories?
  • Jiachen Yu^*, Yufei Zhan^*, Ziheng Wu, Yujiu Yang, Yousong Zhu, Jinqiao Wang
  • Finished during internship in ByteDance.

Internship

• 2025.12-2026.9: Tencent, AI Lab -> Hunyuan Frontier
• 2025.4-2025.11: Bytedance, Data, Global E-commerce
• 2024.10-2025.3: TeleAI
• 2024.6-2024.8: Tencent, TEG

Interests

• Riding, city walk, Running
• Computer Card Games (HearthStone et al.)