Escaping the Diversity Trap in Robotic Manipulation via Anchor-Centric Adaptation
For roboticists deploying VLA models on specific hardware with limited demonstration budgets, this work provides a principled framework to avoid suboptimal data collection strategies.
The paper identifies a 'diversity trap' in robotic manipulation where maximizing coverage with diverse single-shot demonstrations backfires due to estimation noise. It proposes Anchor-Centric Adaptation (ACA), which prioritizes repeated demonstrations at core anchors and selective boundary coverage, achieving significantly higher success rates under a fixed data budget.
While Vision-Language-Action (VLA) models offer broad general capabilities, deploying them on specific hardware requires real-world adaptation to bridge the embodiment gap. Since robot demonstrations are costly, this adaptation must often occur under a strict data budget. In this work, we identify a critical diversity trap: the standard heuristic of "maximizing coverage" by collecting diverse, single-shot demonstrations can be self-defeating due to non-vanishing estimation noise. We formalize this phenomenon as a Coverage--Density Trade-off. By decomposing the policy error into estimation (density) and extrapolation (coverage) terms, we characterize an interior optimal allocation of unique conditions for a fixed budget. Guided by this analysis, we propose Anchor-Centric Adaptation (ACA), a two-stage framework that first stabilizes a policy skeleton through repeated demonstrations at core anchors, then selectively expands coverage to high-risk boundaries via teacher-forced error mining and constrained residual updates. Real-robot experiments validate our trade-off framework and demonstrate that ACA significantly improves task reliability and success rates over standard diverse sampling strategies under the same budget.