Ctrl+K
task_8_seg10: best task-8 result across all tests (q=0.275); needed by the 100-task stock eval
ec1844b verified - T5-DA3-#2
- T5-noDA3-baseline
- bringing_in_wood_20k
- can_meat_20k
- can_meat_30k
- chop_an_onion_20k
- clean_up_your_desk_20k
- da3_eval_samples
- meta5_2026_224
- meta5_2026_224_ext
- norm-stats-100task-2026
- norm-stats-fixed
- norm_stats
- pi_behavior_100t_meta_ft_4task
- pi_behavior_50t_meta_ft_4task
- pibehavior_da3_clean_up_your_desk_40k
- task_8_seg10
- verify
- 407 kB
- 7.1 kB
- 9.13 kB
- 692 Bytes