UOJ-Bench - Repair (Easy): leaderboard

Metric: Pass@1 (%) on the 500 easy repair instances (partial-score submissions): the patched code, changing at most 10 percent of the original, must pass every test, single-turn zero-shot Pass@1 judged by the Universal Online Judge through its internal API (problem statements translated from Chinese to English; maximum response length at least 64k tokens); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.

Top models

#ModelScore
1Gemini 3 Pro (Preview)53.6
2GPT-5 (High)43.8
3GPT-541.4
4Kimi K2 (Thinking)32.2
5DeepSeek V3.2 (Thinking)26.2
6GPT-OSS-120B20.2
7Claude Opus 4.516.6
8Qwen 3 Coder 480B A35B Instruct3.6

Interactive version: theaggregate.ai/benchmark?slug=uoj-bench-repair-easy · How It Works · Data refreshed daily, snapshot 2026-09-29.