[KongchangAI]
ConceptReTerm-9B

ReTerm

A method for a terminal agent to improve itself by learning recovery from its own failed tasks. It adopts a two-stage training paradigm consisting of an SFT cold start followed by online reinforcement learning with ReTerm-GRPO.

Timeline (last 90 days)

Oct 14

ReTerm is a terminal agent that achieves self-improvement by "learning how to recover from its own failed tasks."

Unverified50%
Oct 14

All of ReTerm's code, data, and models have been open-sourced and released in the GitHub repository (twothsnd/ReTerm).

Unverified50%

All Facts (2)

Source Articles