Salesforce: Co-Evolving Harnesses and On-Policy Correction Help Weak Models Catch Up

Saleforce · hf · 2026-09-10

Salesforce research 'Co-Evolving Harnesses and Models' combines harness evolution with localized expert on-policy correction.

The approach helps weaker models improve in settings where imitation fails, without disrupting their native planning style — offering a new lever for boosting small-model performance on agentic tasks.

Original post →

More from coding & agent

coding & agent channel →