Hi Zhang,
Your AutoGen #7487 comment on fanout-weighted integrity checks lined up with something I've been building, so I read through agora-governance. The three-tier separation (judge-only Global Crits, precedent memory, governance layer) is cleaner than most multi-agent governance designs I've seen documented.
How did you land on the precedent weight formula? Empirically tuned from real runs, or closer to first-principles?
Naga
Hi Zhang,
Your AutoGen #7487 comment on fanout-weighted integrity checks lined up with something I've been building, so I read through agora-governance. The three-tier separation (judge-only Global Crits, precedent memory, governance layer) is cleaner than most multi-agent governance designs I've seen documented.
How did you land on the precedent weight formula? Empirically tuned from real runs, or closer to first-principles?
Naga