- Rating
- 2
- Confidence
- 4
- Informativeness
- 3
- Presentation
- 3
- Soundness
- 3
Read review
Paper summary
his paper studies why vision-language-action models fail on concepts absent from robot demonstrations. CrossBind measures target identification separately from task execution. An inherited likelihood readout recovers useful target knowledge across three model families, but its correctness weakly predicts placement in two task-adapted policies. RECALL freezes the action-pretrained backbone and trains an interface and action expert with demonstration target indices. At inference, the inherited readout selects the regions supplied to the expert. A key strength is linking semantic measurement to controlled policy interventions. With policy weights fixed, replacing location-only inputs with selected targets raises success from 21.5% to 43.8% on π0.5, 5.6% to 72.2% on InstructVLA, and 4.9% to 51.4% on X-VLA. However, the main policy results use an earlier data recipe, and each configuration has one training run. Retained semantics can aid control, but the original failure mechanism remains less firmly established.
Critical strengths
- Useful separation of knowledge and behavior. The same-checkpoint analysis tests whether target identification predicts placement. This is more informative than comparing aggregate readout accuracy with task success.
- Clear intervention evidence. Five input conditions share policy weights and evaluation states. Redirected-target and separate object and destination controls confirm sensitivity to the selected referents.
- Consistent gains across three architectures. RECALL reuses inherited scoring without fitting a semantic probe. Each backbone benefits from readout-selected targets over location-only conditioning.
Critical weaknesses
- The strongest shortcut guarantees cover a different recipe. Full-factorial crossing supports the dataset audit, while the main policies use earlier placement-crossed data. This limits how directly those guarantees support the behavioral diagnosis.
- The original failure mechanism remains partly unresolved. The readout uses isolated crops, and RECALL trains a new target-conditioned pathway. These results establish useful retrieval, but also change visual binding demands relative to the original policy.
- Training stability is not established. Each configuration has one training run. Episode-level intervals describe evaluation uncertainty, but do not show whether the gains and weak baseline associations persist across independently trained checkpoints.
Questions for authors
- Do the behavioral findings persist when policies are trained on the full-factorial dataset?
- Does a full-scene readout yield the same weak association with the original policy's placement outcomes?
- How stable are the main within-policy gains and baseline associations across independent training seeds?