LittlePly chess agent shares history representation between policy and value heads
andreisavu · x · 2026-11-08
Part of andreisavu's long thread on the LittlePly chess agent project: the live agent already shares a history representation between policy and value, and the current training experiment includes auxiliary objectives. The author also clarifies that live game outcomes cannot validate an undeployed checkpoint, and that telemetry coverage varies by game — an absent record doesn't prove a component was inactive.
More from Research
- Reviewers reward flaw-free work, users reward useful work — why peer review punishes innovation — wzenus · 2026-11-08
- Math community debates Glazer's proof and OpenAI's H10 paper: slop need not compound — littmath · 2026-11-08
- Stanford paper: more info helps some LLM trading agents, wrecks others — rohanpaul_ai · 2026-11-08
- Neel Nanda pushes back on LRH critique: one steering vector, four norms, four behaviors — NeelNanda5 · 2026-11-08
- AI-generated Community Note fact-checks weight-space programming paper, author clicks No — DimitrisPapail · 2026-11-08
- Anders Sandberg and Danielle Fong debate why simple progress models eventually break down — anderssandberg · 2026-11-08