deciding-robot-skill-has-finished-next-subtask-should-begin CLM-0373For deciding when a robot skill has finished and the next subtask should begin, general-purpose VLMs prompted zero-shot are near useless on long-horizon mobile manipulation (under 4% of transitions detected on BEHAVIOR-1K, including a frontier proprietary model), while a sub-1B student fine-tuned on compact self-explanations condensed from teacher reasoning over annotated demonstrations detects about 90% and lifts closed-loop task success from 0.30 to 0.53 against a 0.63 oracle.
Capability: Turning plans into physical action
Observed on
Hierarchical robot control with a PDDL planner and learned low-level policies; LIBERO-Logic and six BEHAVIOR-1K tasks, 10 training trajectories per task; ground-truth subtask segme.
Sources
- supportsStageGuard: Learning Stage Transitions for Long-Horizon Robot Tasks via Agentic DistillationValidated by: Transition completion, full-trajectory rate, next-subtask accuracy and inference Hz against five prompted VLM baselines and two ROVER variants on LIBERO and BEHAVIOR-1K; ablations against full-teacher-CoT training (0% on BEHAVIOR-1K) and decision-only labels (37.7%); closed-loop success on BEHAVIOR-1K and 20 trials on each of two real arms.. [truncated]
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 VLM agents doing multi-phase navigation and pick-and-place in procedurally generated indoor scenes with moving human obstacles, subtask progress runs well ahead of full success — agents reach the first goal and then stall or loop after the phase transition — so success rate alone hides where the failure is.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 a pretrained vision-language-action policy is fine-tuned sequentially on a stream of real-robot manipulation tasks, the choice of action-normalization statistics is itself a determinant of continual-learning success: statistics that are task-derived, first-task-derived, or recomputed each stage produce forgetting or poor transfer, while statistics estimated once from a small task-independent calibration set covering the embodiment's motion range and then frozen give the best acquisition, retention and task-order stability.Turning plans into physical action · reviewed by AI
- In VLM-driven human-aware robot task planning over scene graphs, the bottleneck is inferring and grounding the symbolic goal, not the planning: once both human and robot goals correctly cover the ground truth, a formal multi-agent planner produces a valid, conflict-free joint plan in 99.6% of trials, while end-to-end success with VLM-inferred human goals is 69.7%.Planning a sequence of actions before acting · reviewed by AI
- When a calculator-using model is trained with reinforcement learning on verifiable final-answer rewards for an arithmetic search task (Countdown), the gain lands almost entirely at low k — pass@1 rises sharply while pass@16 moves little or falls — because the update can only reinforce correct trajectories the starting policy already samples, and prompt groups with no correct sample supply no gradient.Using the tools it is given · 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: medium. Falsifier as drafted: A prompted general VLM reaching comparable transition-completion rates on long-horizon mobile manipulation, or the self-explanation step giving no advantage over direct full-CoT or decision-only fine-tuning on a comparable benchmark.