Simon Yu ✈️ EMNLP2026(@simon_ycl):Recursive self-improvement needs agents that can build their own training environments. But researchers still part of the loop, making it hard to scale. We introduce AutoEnvScaling, which turns Data Flywheel into a terminal task: a proposer agent builds complete Harbor tasks and sandbox. It builds just like human researchers. This enables self-play RL for terminal agents: one model improves as both builder and solver, without human-in-the-loop.

Simon Yu ✈️ EMNLP2026

@simon_ycl

Building Self-improvement system (Shepherd, SPADE) | 3rd Year PhD, supervised by @shi_weiyan | intern in @MSFTResearch | Prev Student @EdinburghNLP

가입 September 2015

979 팔로잉 중 922 팬

Simon Yu ✈️ EMNLP2026@simon_ycl

2026.10.14 13:03

Recursive self-improvement needs agents that can build their own training environments. But researchers still part of the loop, making it hard to scale. We introduce AutoEnvScaling, which turns Data Flywheel into a terminal task: a proposer agent builds complete Harbor tasks and sandbox. It builds just like human researchers. This enables self-play RL for terminal agents: one model improves as both builder and solver, without human-in-the-loop.