Four-Factor Integrity Breakdown & Primary-Source Evidence
Access Model: Public Plaintext
"The hidden tests ship inside the task directories (Harbor tasks carry their tests), so this is a public benchmark: do not train on it."
"Given a codebase containing many concurrent bugs and no information about their nature or location, an agent must autonomously discover and fix as many bugs as possible."
Originality
Weight 25%
4/10
Reuse Profile: Medium Reuse (25-75%)
"SWE-sweep is constructed from open-source repositories. For each repository, we collect issue-pull request pairs, then identify a single commit where the maximum number of bugs are present at the same time."
Freshness
Weight 20%
6/10
Initial Release: 2026-10Latest Refresh: 2026-11Rolling Updates: No
"100 repositories · 4.1k bugs · Updated November 2, 2026"
Integrity: Independent Offline (Leak-Verified)Sandbox: offline isolated
"upstream runs the agent container with --network=none"
Benchmark maintainers can embed this live SVG badge in GitHub READMEs or documentation:
[](https://benchaudit.org/benchmarks/swe-sweep/)