Jonathan @RLC

@lightetal

I’m a PhD researcher @MSFTResearch @Caltech @RPI, prev @NECLabsAmerica @AsariAILabs working on LLM-agents, reasoning, RL, test-time scaling, and code generation

San Francisco, CA

jonathanmli.github.io

Joined June 2023

Pinned
Jonathan @RLC
@lightetal
Feb 26
Post-training LLMs is like mixing a cocktail: Too much easy data → no learning Too much hard data → instability Wrong balance → collapse And today, we mix it by hand. What if the data mixture could be learned instead of hand-tuned? arxiv.org/abs/2602.20532 🧵👇
Jonathan @RLC
@lightetal
Aug 15
Really excited for my second @RL_Conference ! Looking forward to chatting about RL for increasingly capable agents, self-improvement, curriculum/continual learning, and inference-time search. Would love to meet others thinking about these problems. Say hi at #RLC2026!
Jonathan @RLC
@lightetal
Aug 12
Super curious about how this enables collective self-improvement, much like how human society functions today!
Yisong Yue
@yisongyue
Aug 5
Knowledge flywheels are a new scaling dimension for self-improving AI.
Article
Knowledge Flywheels
A new scaling dimension is emerging Today, we scale models through better data, architectures, and compute. We scale agents through better tools, search, verification, and harnesses. Tomorrow, we will...
Jonathan @RLC
@lightetal
Aug 12
Excited to share that I’m at @MSFTResearch in Montreal until October, working on coding agents and new techniques for agentic training! Hit me up if you’re in the area and want to chat about RL, agents, code generation, post-training, or just grab a drink!
Jonathan @RLC
@lightetal
Aug 12
Amazing work by the team at Asari AI!
Asari AI
@AsariAILabs
Jul 29
Our self-improving agents optimized the full @vllm_project inference stack, with up to 16% more throughput and interactivity for @deepseek_ai v4 Pro and @Zai_org GLM 5.2 on B200s (no MTP). Every change was verified and our agents got better and faster at it with each iteration.
00:00