# SWE-sweep Contamination Audit

> SWE-sweep scores 3.7/10 (Tier C, High Risk) on BenchAudit as of November 2026, reflecting publicly downloadable plaintext test files and isolated offline third-party grading.

- **Canonical URL:** https://benchaudit.org/benchmarks/swe-sweep/
- **Benchmark ID:** `swe-sweep`
- **Hub:** Meta Superintelligence Labs
- **Category:** Coding
- **Domains:** Coding
- **Skills:** Coding, Agentic, Tool Use, Reasoning, Instruction Following
- **BenchAudit Score:** 3.7/10 (Tier C - High Risk)
- **Last Audited:** 2026-11-19 (Ref 2026-11)

## Factor Scores

| Factor | Score (/10) | Weight |
| --- | ---: | ---: |
| Privacy | 0 | 40% |
| Originality | 4 | 25% |
| Freshness | 6 | 20% |
| Grading | 10 | 15% |

## Primary-Source Evidence

### Privacy
> "The hidden tests ship inside the task directories (Harbor tasks carry their tests), so this is a public benchmark: do not train on it."
- Source: https://raw.githubusercontent.com/facebookresearch/swe-sweep/main/tasks/README.md

### Originality
> "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."
- Source: https://swesweep.com/

### Freshness
> "100 repositories · 4.1k bugs · Updated November 2, 2026"
- Source: https://swesweep.com/

### Grading
> "upstream runs the agent container with --network=none"
- Source: https://raw.githubusercontent.com/facebookresearch/swe-sweep/main/tasks/swe-sweep__agda/task.toml
