Files
nagi-bench/outputs
Jesse ZhangandClaude Opus 4.8 63570b1e6e Add Kimi K3 (Kimi Code, Max) — mythos, pelican, turf-war 蚕食
Turf-war agent registered (policy stays private on the referee side).
Diagnosis: legal + fuel-safe but weak, loses ~3:1 to baseline (318:1045).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-17 01:22:54 +01:00
..