Curated subsets of nebius/SWE-rebench-openhands-trajectories for fine-tuning evaluation experiments.
DimAblation-no_action_diversity
500
Single dimension removed: action_diversity
0.9456
100%
100%
Ablation study
DimAblation-no_observation_noise
500
Single dimension removed: observation_noise
0.9299
100%
100%
Ablation study
DimAblation-no_outcome_success… See the full description on the dataset page:
https://huggingface.co/datasets/davongluck/swe-trajectory-subsets.