flâneur

Honey, I Looked at the Data of a Frontier Benchmark and Found some Issues: Porting Agents' Last Exam Linux CLI subset to Verifiers v1 – sankalp's blog

sankalp.bearblog.dev · saved by 3 readers

meta_description: Porting Agents’ Last Exam to verifiers uncovered hidden output requirements, answer leakage, and broken tasks. Here’s how that led to ALE-Gold.

saved by