Category

Safety & evaluation

Testing what AI gets wrong: research on evaluation, benchmarks, robustness and safety, and on failures that a single accuracy score hides.

Follow Safety & evaluation
IIIOpen sourcePreprint

Two small training edits let a sandbagging model slip past a text audit

A preprint shows that a detector which nearly always catches a model that underperforms on purpose can be pushed to chance once the attacker adapts. Probes that read the model's internals mostly picked up fine-tuning in general, not the hidden behaviour.