MemT-1.0
🧠 My role: Technical Advisor, leading the reinforcement-learning stage
100K+ Downloads in Week 1
MemT-1.0 makes memory an intrinsic part of the model’s computation rather than an external retrieval step: a dynamic local memory and a trainable hyper-memory consolidate past interactions into a fixed-size persistent memory state, so the model can draw on prior sessions without re-reading the original text at inference time.
Key results:
- ⚡ At a 128K history length, Nano achieves about a 1.86× end-to-end inference speedup and a 66× reduction in persistent storage compared with a full-context KV cache
- 📈 Outperforms parametric-memory baselines on average, stays competitive with retrieval-augmented methods, and largely preserves general capabilities
- 📦 Open-sourced in Oct. 2026 with two models (Nano 36B-A3B, Flash 288B-A17B) and a 66-page technical report; 100K+ downloads within one week of release
Authors
Haotian Chen
(he/him)
Research Assistant Professor · LLM Post-Training / AI4AI
I am a Research Assistant Professor at the School of Artificial Intelligence, Shanghai Jiao Tong University, where I work with Prof. Junchi Yan at RethinkLab. I study how to build AI systems that can automate long-horizon, effort-intensive, and creativity-demanding tasks such as research, engineering, and development. My current work focuses on LLM post-training for autonomous agents, agent memory and continual learning, and AI4Research / AI4AI. Before joining SJTU, I received my PhD in Data Science from Fudan University, advised by Prof. Xiangdong Zhou, and completed postdoctoral research at Tsinghua University (THUNLP), working with Prof. Zhiyuan Liu and Prof. Maosong Sun. I was also a research intern at the Machine Learning Research Group of Microsoft Research Asia, mentored by Xiao Yang, and at the Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, working with Prof. Yang Yu.