Tsinghua and Moonshot AI Launch TCSExam: GPT-6 Astra Proves 70% of Theorems but Solves No Open Conjectures
Tsinghua University and Moonshot AI released TCSExam, a 100-task expert-reviewed benchmark spanning Lean theorem proving, solver engineering, record-setting constructions, and 30 open conjectures, with every submission verified by the Lean kernel or a task checker. GPT-6 Astra fully proved 70.0% of theorems without supplied lemmas, versus 38.3% for Kimi K3.1 Pro and at most 11.7% for six other agents, yet the same model solved none of the 30 open TCS conjectures; the few accepted conjecture submissions merely adapted counterexamples and proofs OpenAI had already published.
Related Entities
Full Report
Rohan Paul (@rohanpaul_ai) · 2026-11-08 18
New Tsinghua and Moonshot AI paper finds that GPT-6 Astra proves 70.0% of research-level computer science theorems in Lean, yet the same model resolves 0 of 30 open conjectures in theoretical computer science (TCS).
Theoretical computer science is a natural test for AI research, because proofs can be machine-checked and records can be beaten. Existing benchmarks cover only slices, like proofs or code.
TCSExam has 100 expert-reviewed tasks in 4 tracks: Lean theorem proving, solver engineering, record-setting constructions, and 30 open conjectures. The Lean kernel or a task checker verifies every submission.
Without supplied lemmas, GPT-6 Astra fully proved theorems in 70.0% of attempts, Kimi K3.1 Pro in 38.3%, and the other 6 agents in at most 11.7%. The few accepted conjecture submissions adapted counterexamples and proofs OpenAI had already published.
