https://zijunsong.github.ioResearch notes on agents and post-traininghttps://zijunsong.github.io/posts/harnessopd.htmlhttps://zijunsong.github.io/posts/harnessopd.htmlA stronger search-agent harness can improve decisions, but the gains may vanish once those components are removed. HarnessOPD studies how to keep part of that advantage in the model parameters.https://zijunsong.github.io/posts/rg-cto.htmlhttps://zijunsong.github.io/posts/rg-cto.htmlFailed traces should not be reused wholesale. RG-CTO estimates the reliability of negative experience before deciding how strongly it may shape generation.