Tsinghua University and Moonshot AI release TCSExam benchmark: GPT-6 Astra proves 70.0% of research-level theorems but cannot solve any open conjectures.
Tsinghua University and Moonshot AI have released the TCSExam benchmark, which evaluates large language models' capabilities across the entire spectrum of theoretical computer science research using 100 expert-reviewed tasks. The test covers Lean theorem proving, solver engineering, constructive records, and 30 open conjectures, with all submissions verified by the Lean kernel or task checkers.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open AIHOT original