Benchmark
EvasionBench
What is EvasionBench?
A benchmark of 50 diverse task-policy pairs where task completion requires an operation prohibited by a runtime monitor.
- Released
- 2026-09-24
- Evaluates
- General AI, Language & Knowledge, cs.CR
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- ai-name-audit-deferred
- Reported scores
- 0
Source provenance
- Original evidence https://arxiv.org/abs/2609.30217
EvasionBench paper
No reported scores are on record for this benchmark yet.