anote-ai/

codebench-results

Experimental results from the CodeBench evaluation framework. Two dataset configs are included.

	Configs




	


	h4_security — H4 Security-Adjusted Reliability Experiment

240 rollouts across 3 agents × 10 tasks × 8 rollouts. Tasks split into 5 standard algorithmic tasks and 5 security-sensitive tasks (eval, exec, shell, yaml, dynamic import).

Column Description

task_id Task identifier (algo-* or sec-*)

agent_name anote-code… See the full description on the dataset page: https://huggingface.co/datasets/anote-ai/codebench-results.