AgentNativeResearchLab/arc-agi3-cc-opus4.8-ls20-demo
Updated • 314
Full record per harness×model×game: trajectories, recordings, replays + the ARA the agent built live. Same game across models = comparison unit.
Note ls20 ablation / showcase demo
Note June archive baseline — pre-ARA-method recording, old format
Note June archive baseline — pre-ARA-method recording, old format
Note June archive baseline — pre-ARA-method recording, old format
Note June archive baseline — pre-ARA-method recording, old format
Note June archive baseline — pre-ARA-method recording, old format
Note June archive baseline — pre-ARA-method recording, old format