v1.3.2

Changelog

Reconstruct the reward system and align with CoMLRL 1.3.2.

ClassEval

IAC and MAGRPO ested on H200 (33 and 48 h), MAAC need B200 (estimated time 50 h)

v1.0.0

This version provides excellent support for training scenarios involving two agents collaborating.

Model links:

https://huggingface.co/OpenMLRL/CE_2agents_TAKE_JOB_2t_code_feedback_0p
https://huggingface.co/OpenMLRL/CE_2agents_TAKE_JOB_2t_code_feedback_1p