rlundqvist/vea-rm-stress-tests download history

rlundqvist/vea-rm-stress-tests is a text classification dataset on the Hugging Face Hub. In the last 30 days it was downloaded 26 times (5 in the last 7 days), and 86 times in total. It ranks #306,850 among datasets by monthly downloads.

RM VEA Stress Tests — does the RM penalize eval-awareness even when it's aligned? 104 matched items (8 task domains). Each item is a competent model response in four forms that differ only by one inserted chain-of-thought clause: field what it adds base no eval-awareness at

Open rlundqvist/vea-rm-stress-tests on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.