← 研踪 PaperTrailarXiv:2611.22302 · v1
Where Do Human-to-Robot Transfer Gains Come From?
首次提交 2026-11-17 14:59
版本更新 2026-11-17 14:59
雷达观测 2026-11-18 19:43
摘要
Human data can support robot policy pretraining, but gains attributed to a data recipe may also reflect subset size, training variability, or downstream adaptation. In an HPT implementation on a 20-task MetaWorld suite with three training blocks, a constrained random 5,000-recording subset matches most of the task-informed subset's gain over the 40,000-recording pool at 20,000 updates (3.69 of 4.29 percentage points), leaving equal-size selection differences of 0.60 points at 5,000 recordings and 0.45 points at 1,250 recordings, with wide paired intervals. Across 12 comparisons, task orderings repeat across seeds for several contrasts, but a fixed predictor built from the other two seeds trails the mean or difficulty reference in at least one held-out fold for every comparison. Neither mechanism probe shows its predicted positive interaction: task-slot mismatch yields an interaction of points after adaptation, and in a six-task observation experiment the observation-by-pretraining interaction changes sign between seeds ( and points) while full-state pretraining gains are positive in both ( and points). This retrospective study shows how equal-size and capacity controls, seed-level reporting, and targeted interventions change the interpretation of apparent transfer gains.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-17 14:59 · 最新观测版本
Where Do Human-to-Robot Transfer Gains Come From?
Human data can support robot policy pretraining, but gains attributed to a data recipe may also reflect subset size, training variability, or downstream adaptation. In an HPT implementation on a 20-task MetaWorld suite with three training blocks, a constrained random 5,000-recording subset matches most of the task-informed subset's gain over the 40,000-recording pool at 20,000 updates (3.69 of 4.29 percentage points), leaving equal-size selection differences of 0.60 points at 5,000 recordings and 0.45 points at 1,250 recordings, with wide paired intervals. Across 12 comparisons, task orderings repeat across seeds for several contrasts, but a fixed predictor built from the other two seeds trails the mean or difficulty reference in at least one held-out fold for every comparison. Neither mechanism probe shows its predicted positive interaction: task-slot mismatch yields an interaction of points after adaptation, and in a six-task observation experiment the observation-by-pretraining interaction changes sign between seeds ( and points) while full-state pretraining gains are positive in both ( and points). This retrospective study shows how equal-size and capacity controls, seed-level reporting, and targeted interventions change the interpretation of apparent transfer gains.
查看此版本解读 →查看此版本原文 ↗