TCSExam — 2 Technical Articles

Nov 10, 2026 · Artificial Intelligence

Kimi Breaks Graph Theory Record in Yao Qizhi & Yang Zhilin's New TCSExam Benchmark

A new benchmark co-authored by Turing laureate Yao Qizhi and Kimi founder Yang Zhilin tests LLMs on 100 theoretical CS tasks; Kimi K3 constructs a 630-edge graph beating the prior 629-edge record, while GPT-6 Astra improves a quantum error correction code's distance from 5 to 6.

GPT-6 AstraKimiLLM benchmarking

0 likes · 9 min read

Kimi Breaks Graph Theory Record in Yao Qizhi & Yang Zhilin's New TCSExam Benchmark

PaperAgent

Nov 9, 2026 · Artificial Intelligence

TCSExam Reveals LLMs' Differentiated Research Skills: Theorem Proving Strong, Autonomy Weak

The TCSExam benchmark evaluates frontier LLMs on four theoretical computer science research tasks—theorem proving, solver engineering, mathematical construction, and open conjectures—using verifiable Lean 4 proofs and executable checkers; GPT-6 Astra leads at 70% theorem-proving success, Kimi K3.1 Pro follows at 38%, but models show divergent strengths across tasks and still require human-provided lemmas for complex proofs.

AI researchLLM benchmarkLean 4

0 likes · 10 min read

TCSExam Reveals LLMs' Differentiated Research Skills: Theorem Proving Strong, Autonomy Weak