TCSExam: GPT-6 Astra Proves 70% of Theorems but Solves Zero Open Conjectures

A Tsinghua and Moonshot AI paper introduces TCSExam, where GPT-6 Astra topped the full benchmark with 71.9 points and proved 70% of research-level TCS theorems in Lean, yet failed to solve any of 30 open conjectures. Opus 5.5 led the compact version.

2026-11-09 ~ 2026-11-09 · 2 related posts

• Tsinghua & Moonshot's TCSExam: GPT-6 Astra proves 70% of theorems, solves 0 open conjectures — rohanpaul_ai · 2026-11-09
• TCSExam benchmark: GPT-6 Astra tops Full at 71.9, Opus 5.5 leads Lite at 72.3 — rohanpaul_ai · 2026-11-09