JingweiNi/iclr-2action-controller-chains download history

JingweiNi/iclr-2action-controller-chains is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 138 times (34 in the last 7 days), and 162 times in total. It ranks #90,436 among datasets by monthly downloads.

ICLR 2-action controller — rollout chains Chain-of-thought rollouts with counterfactual stopping probes, for a single-chain {continue, terminate} controller on Qwen3-1.7B (native thinking). Each JSONL line is one chain: field meaning qkey, rep, gold question id, chain index

Open JingweiNi/iclr-2action-controller-chains on Hugging Face

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