AI Agent · Claude · AI Safety Newsletter
MLSN #23: AIs Leak Their Values Into Factual Questions
Compiled by KHAO Editorial — aggregated from 1 source. See llms.txt for citation guidance.
◌ Single Source
TLDR: A new benchmark measures how AIs improve at tasks after clear feedback and iteration.
Key facts
- AIs learning tasks over time has significant real-world consequences even without RSI: AI agents have already launched autonomous cyberattacks, relying significantly on iteration over weeks
- EdgeBench comprises 134 tasks, with each task requiring an average of 57.2 hours of human expert work to complete
- AIs completing these tasks are scored on a rubric from 0% to 100%, and they get automated feedback on their proposed solutions over time, allowing for improvement and learning
- However, the CoTs of highly capable models like Claude Fable 5 are often protected by exfiltration monitors that prevent such jailbreaks
Summary
Existing AI capability benchmarks primarily measure whether AIs can complete tasks on their first try. EdgeBench comprises 134 tasks, with each task requiring an average of 57.2 hours of human expert work to complete. Beyond simply getting higher scores overall, AIs have been getting faster at learning EdgeBench tasks. While previous capability benchmarks often measure narrow software engineering ability, EdgeBench measures AI agents on a diverse array of tasks that current AI agents cannot solve. EdgeBench measures the ability to iterate on problems over many attempts and improve solutions based on feedback.