Codesota · Computer Code · Bug Detection · Bugs2FixTasks/Computer Code/Bug Detection
Bug Detection · benchmark dataset · 2019 · EN

Bugs2Fix: Learning to Rewrite Buggy Code.

Bug detection and repair benchmark with ~2.4M Java methods mined from GitHub commits labeled as bug fixes. Used widely to evaluate LLM bug detection capabilities. Primary metric is Accuracy (correct bug classification).

Paper ↗Submit a result ↵
§ 01 · Leaderboard

Best published scores.

6 results indexed across 1 metric. Shaded row marks current SOTA; ties broken by submission date.

Primary
accuracy · higher is better
accuracy· primary
6 rows
#ModelOrgSubmittedPaper / codeaccuracy
01GPT-4oAPIOpenAIMar 2026arxiv78.60
02Qwen2.5-Coder 32BOSSAlibabaSep 2024Qwen2.5-Coder Technical Report · code76.80
03DeepSeek-Coder-V2-InstructOSSDeepSeekJun 2024DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source… · code75.30
04CodeT5+OSSSalesforceMay 2023CodeT5+: Open Code Large Language Models for Code Unders… · code68.20
05UniXcoderOSSMicrosoftMar 2022UniXcoder: Unified Cross-Modal Pre-Training for Code Rep… · code66.40
06CodeBERTOSSMicrosoftFeb 2020CodeBERT: A Pre-Trained Model for Programming and Natura… · code62.50
Fig 2 · Rows sorted by score within each metric. Shaded row marks SOTA. Dates reflect model or paper release where available, otherwise the date Codesota accessed the source.
§ 03 · Progress

6 steps
of state of the art.

Each row below marks a model that broke the previous record on accuracy. Intermediate submissions are kept in the leaderboard above; only SOTA-setting entries are re-listed here.

Higher scores win. Each subsequent entry improved upon the previous best.

SOTA line · accuracy
  1. Feb 19, 2020CodeBERTMicrosoft62.50
  2. Mar 7, 2022UniXcoderMicrosoft66.40
  3. May 13, 2023CodeT5+Salesforce68.20
  4. Jun 17, 2024DeepSeek-Coder-V2-InstructDeepSeek75.30
  5. Sep 19, 2024Qwen2.5-Coder 32BAlibaba76.80
  6. Mar 27, 2026GPT-4oOpenAI78.60
Fig 3 · SOTA-setting models only. 6 entries span Feb 2020 → Mar 2026.
§ 04 · Literature

5 papers
tied to this benchmark.

Every paper below corresponds to at least one row in the leaderboard above. Click through for the arXiv preprint and, when available, the reference implementation.

§ 06 · Contribute

Have a score that beats
this table?

Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.

Submit a result ↵Read submission guide
What a submission needs
  • 01A public checkpoint or API endpoint
  • 02A reproduction script with frozen commit + seed
  • 03Declared evaluation environment (Python, deps)
  • 04One row per metric declared by this dataset
  • 05A contact so we can follow up on discrepancies