PreAct-Bench
Explore score history ↗PreActBench is a benchmark for predictive monitoring in LLMs, consisting of 1,000 paired ethical and unethical action trajectories across five domains. It evaluates whether models can infer if a partial trajectory will culminate in unethical action, using the Prefix Foresight F1 metric.
- Released
- 2026-06-03
- Institutions
- King's College London · National University of Singapore · Southern University of Science and Technology
- Evaluates
- General AI, Language & Knowledge, cs.LG
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- unreviewed
PreAct-Bench paper
No reported scores are on record for this benchmark yet.