Research runs (newest first). Each run is a self-contained snapshot of one execution of the multi-agent pipeline.
2026-05-23 latest
claude-sonnet-4-6 · Template 1 + Template 2 v1 (5 命题) · tier=sanity-check (pre-v1.4) · ~75 agents · ~$8–12
P1 PASS (trivial) · P2 KILL · P3 KILL · P4 KILL · P5 PASS (position paper)