manipulation-policies-diffusion-policy-vla-backbones-trained-few-demonstrations-decomposing CLM-0608observationsingle paperReviewed by AI
For manipulation policies (diffusion-policy and VLA backbones) trained on few demonstrations, decomposing trajectories implicitly via a Mixture-of-Experts gate and storing the per-segment gating profiles as retrievable memory, then blending retrieved profiles into the current gating distribution at inference, raises success most on unseen object-receptacle recombinations rather than on in-distribution tasks.
Capability: Turning plans into physical action
Observed on
LIBERO suites, Push-T/BlockPush/Kitchen, and five real UR5e tasks with ~50 demonstrations each; gains are small in-distribution (e.g. 96.8% to 98.0% over pi-0.5) and larger on cros.
Sources
- Validated by: Task success rate against the unmodified backbone and prior memory-augmented VLAs on LIBERO (four suites), three DP control benchmarks, and 40 real-robot rollouts per task; ablation separates segmentation alone from segmentation plus memory.. Measured on one benchmark family plus one real robot; ablations report segmentation-only versus segmentation-plus-memory, and a cross-suite zero-shot table. No matched-parameter control (SkillMemo adds parameters), no random-retrieval control, and real-world results are 40 trials per task from one lab.
Either way, bring evidence: a link, or what you saw.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- For memory-augmented VLA manipulation policies that must actively move the viewpoint to find task-relevant evidence, writing one memory frame at each fixed action-chunk boundary generalizes better than event-triggered writing (on predicted subtask change or on gripper/head/torso motion), and extra memory capacity helps only under the regular schedule — event-triggered policies degrade further out of distribution because mistimed triggers omit informative views and control jitter fills memory with redundant frames.Turning plans into physical action · reviewed by AI
- When adapting a pretrained VLA manipulation policy with only a handful of target demonstrations, retrieving prior demonstrations at the level of LLM-decomposed subtasks (then reranking by behavior-feature similarity) raises success more than retrieval by visual, optical-flow, or state-action similarity, and the margin is concentrated on long-horizon composite tasks rather than single-skill pick-and-place.Turning plans into physical action · reviewed by AI
- When 4B-8B LLM agents are trained with plain group-based RL (GRPO) to emit variable-length action chunks in text environments, the policy degenerates — collapsing to one action per round on one backbone and over-committing to 5-6-action chunks with lower success on another — because a sparse terminal reward carries no information about where chunk boundaries belong; supplying boundary targets from subskills segmented out of successful trajectories yields an intermediate 3-4-action regime with both higher success and fewer decision rounds.Completing tasks that take many steps or many hours · reviewed by AI
- For post-training a pretrained vision-language-action manipulation policy under a fixed per-round frame budget, collecting handheld demonstrations only at states where the human action chunk diverges from the policy's sampled action distribution, and weighting them by a learned task-progress advantage, raises task progress more than collecting the same number of frames of ordinary demonstrations and fine-tuning on them.Turning plans into physical action · reviewed by AI
- For manipulation policies whose large VLA backbone cannot run within the control period, letting the slow model propose action chunks from a stale observation while a small RL-trained edit policy adjusts and Q-selects among those chunks using the latest observation recovers more performance than applying RL on top of asynchronous imitation-learning schemes alone, and the gap widens as inference delay or environment speed grows.Turning plans into physical action · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: low. Falsifier as drafted: An ablation under matched parameters and inference budget showing the memory component adds nothing beyond MoE segmentation on unseen object-receptacle combinations, or that the cross-suite gain disappears when the retrieval similarity is randomized.