FormalTCS - Autoformalization: leaderboard

Metric: BEq+ equivalence rate (%; natural-language theorem translated to a Lean 4 statement, counted when bidirectional proof search shows it equivalent to the reference statement; 8 sampled candidates per instance at temperature 0.6; 143 instances from theoretical computer science papers accepted at STOC, FOCS, SODA, COLT and similar venues in 2025-2026, each stage given the human-annotated input; each model runs in its own vendor harness). Source: arxiv.org. Saturation forecast: Around 2029. 8 models tracked.

Top models

#ModelScore
1Claude Opus 5 (Claude Code)11.2
2GPT-5.6 Sol (Codex)10.5
3Claude Sonnet 5 (Claude Code)9.1
4Claude Haiku 4.5 (Claude Code)2.1

Interactive version: theaggregate.ai/benchmark?slug=formaltcs-autoformalization · How It Works · Data refreshed daily, snapshot 2026-09-26.