ProgramBench hidden-test pass rate after episode 1
Program-reconstruction hidden-test pass rate after the first of five sequential long-context episodes.
1 models
Higher is better.
About this benchmark
- Metric
- Hidden-test pass rate after episode 1
- Tasks
- 166 golden program-reconstruction tasks
- Authors
- Yang et al. (2026)
Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.