Papers

Trending research and the full catalog - each paper linked to the benchmarks, methods, and models it introduces.

Filtered by domain: safety-and-groundingClear

Hardening Agent Benchmarks with Adversarial Hacker-Fixer Loops

8 Jun 2026

Agent benchmarks score submissions with outcome verifiers that are typically hand-written and brittle, leaving them open to reward hacking. We audit 1,968 tasks across five terminal-agent benchmarks and find 323 (16%) hackable by frontier models given only the task description.

Agents Reinforcement Learning Safety and Grounding

160.0/h

Filtering here covers the 2,000 most recent papers, as much as one page can hold in memory. See the full index of 22,663 papers.