coolguy 3b6da3bcbc Record R1 (discard_only opponent) result + R2 direction ...

R1 결론: trap 깨지지 않음. 자기참조 회로는 충분조건이 아니라 trap 강화
요인일 뿐. opponent를 fixed external로 바꾸자 모델이 zero-pit으로
collapse — 핵심 결함은 model 자체의 credit assignment 실패임이 확정.

R2 방향 = curriculum (작은 게임 → 큰 게임)을 후속 후보로 등록.

2026-05-10 19:04:12 +09:00

..

archive

Refresh optimization plans

2026-05-07 22:15:09 +09:00

deep-cfr-selectivity.md

Record R1 (discard_only opponent) result + R2 direction

2026-05-10 19:04:12 +09:00

librarian.md

Close librarian: full archive promote-survey + parallel dispatch

2026-05-08 02:33:51 +09:00

model_size_experiment.md

Rename bot family, curate analyze plots, tier evaluation cadence

2026-05-10 15:32:55 +09:00

option_b_interleaved_traversal.md

Rename bot family, curate analyze plots, tier evaluation cadence

2026-05-10 15:32:55 +09:00

torch_compile.md

Rename bot family, curate analyze plots, tier evaluation cadence

2026-05-10 15:32:55 +09:00