← Wednesday, September 30, 2026Models & trainingPaper

ROSS selectively reuses useful behavior from historical model rollouts

huggingface.co by Zhiwei Zhang et al. Editor score 8/10

ROSS keeps historical trajectories as context but applies training loss only to selected continuations; on Qwen3.6-35B-A3B, it raises the reported MOPD average from 58.40% to 62.20% and SWE-bench Verified from 64.20% to 68.40%. It offers an evidence-backed way to reuse on-policy experience without imitating its mistakes.

Read on huggingface.co →

Headline, summary, brief and diagram are written by a model from the source's own text; the link, byline and numbers come from the source. Read the original before relying on a claim.