AI Safety Evaluation Datasets Datasets and viewer resources for studying evaluation awareness, instrumental behaviour, sabotage, and monitoring in AI agents. aisa-group/ResearchArena-Trajectories Updated Sep 23 • 1.69k • 5 aisa-group/instrumental-choices-agent-traces Viewer • Updated 3 days ago • 1.68k • 688 • 1 aisa-group/EvalAwareBench Viewer • Updated Sep 29 • 103k • 338 • 7 aisa-group/ResearchArenaMax-Trajectories Updated Aug 13 • 55 • 3
Agent Trajectories Agent trajectories from the AI Safety and Alignment Group’s benchmarks, released for analysing agent behaviour, capabilities, and safety. aisa-group/PostTrainBench-Trajectories Updated about 17 hours ago • 130k • 17 aisa-group/InferenceBench-Trajectories Updated Jun 21 • 849 • 4 aisa-group/ResearchArenaMax-Trajectories Updated Aug 13 • 55 • 3 aisa-group/instrumental-choices-agent-traces Viewer • Updated 3 days ago • 1.68k • 688 • 1
AI Safety Evaluation Datasets Datasets and viewer resources for studying evaluation awareness, instrumental behaviour, sabotage, and monitoring in AI agents. aisa-group/ResearchArena-Trajectories Updated Sep 23 • 1.69k • 5 aisa-group/instrumental-choices-agent-traces Viewer • Updated 3 days ago • 1.68k • 688 • 1 aisa-group/EvalAwareBench Viewer • Updated Sep 29 • 103k • 338 • 7 aisa-group/ResearchArenaMax-Trajectories Updated Aug 13 • 55 • 3
Agent Trajectories Agent trajectories from the AI Safety and Alignment Group’s benchmarks, released for analysing agent behaviour, capabilities, and safety. aisa-group/PostTrainBench-Trajectories Updated about 17 hours ago • 130k • 17 aisa-group/InferenceBench-Trajectories Updated Jun 21 • 849 • 4 aisa-group/ResearchArenaMax-Trajectories Updated Aug 13 • 55 • 3 aisa-group/instrumental-choices-agent-traces Viewer • Updated 3 days ago • 1.68k • 688 • 1
