The paper examines latent action models that encode transitions between video frames using algebraic constraints such as additive composition and antisymmetric reversal. It demonstrates that these algebraic consistency conditions do not reliably certify temporal structure, as unconstrained models can achieve similar error reductions and constrained models still outperform unconstrained ones even after temporal pairings are destroyed. The authors find that preserving temporal pairing offers no consistent advantage on downstream tasks and that a direct repair objective yields only marginal improvement, recommending a more rigorous validation protocol.
arXiv:2606. 19408v1 Announce Type: new Abstract: Latent actions provide a compact interface between action-free video and downstream decision-making, yet existing Latent Action Models (LAMs) force every transition through a fixed-capacity bottleneck.
arXiv:2609.39563v1 Announce Type: new Abstract: Existing video language models encode sampled RGB frames independently, so a long video must either exhaust the token budget or drop the changes betwee...
ActionPiece rethinks how actions are tokenized for autoregressive vision‑language‑action models by introducing physical rank consistency (PRC) to evaluate relational fidelity of reconstructed actions. The method jointly supervises representation learning and quantization to preserve local physical distance rankings, improving both PRC and policy success. Experiments on LIBERO, LIBERO‑Plus, SimplerEnv, and VLA‑Arena show significant gains over baseline tokenizers.
ARC‑Bench is a new benchmark that tests whether frozen JEPA‑style latent world models can correctly rank candidate actions by latent distance. The study finds that the assumption of latent rankability fails dramatically in both navigation and manipulation tasks, with the top‑scored actions often being suboptimal. Closed‑loop replanning masks this defect, but reducing replanning frequency reveals the underlying ranking failures.
arXiv:2609.37297v1 Announce Type: new Abstract: Cross-skeleton motion generation trains generative models to carry action structure and motion intention from one body to another. Yet a target motion...
arXiv:2608. 08982v1 Announce Type: new Abstract: Interactive video world models generate rollouts autoregressively under an action stream, yet they are trained and evaluated almost exclusively on factual prediction.
The study examines how different vision‑language‑action (VLA) policies execute a manipulation task by comparing the geometry of their end‑effectors across 15,000 closed‑loop LIBERO rollouts. By pairing 3,600 configuration‑matched policy executions, the authors find that when both policies succeed, their end‑effector trajectories are much closer (median DTW distance 0.0120 m) than when only one succeeds (0.0380 m), a pattern consistent across all tasks, policy pairs, and nine representations. Even successful executions remain as far from same‑task demonstrations as the demonstrations are from each other, indicating that task‑associated geometry, rather than training data overlap, drives these differences.
arXiv:2607. 27138v1 Announce Type: cross Abstract: Vision-language-action (VLA) models remain constrained by scarce action-labeled robot data, whereas action-free videos offer abundant observations of physical change.
The paper introduces Action Forcing, a method that transforms ordinary unlabeled video into action‑supervised training data by extracting egomotion bases through principal component analysis of pixel displacements. This approach yields grounded throttle–yaw control signals without requiring instrumented platforms or manual annotation, and it trains a high‑capacity video model while preventing pixel‑level overfitting via an online latent critic. The authors also critique standard video generation metrics and propose a reference‑free evaluation that measures controllability, plausibility, conjuring, and geometric integrity, showing that their model can reverse, scale, and compose actions despite limited reverse‑action data.
RotVLA introduces a Vision‑Language‑Action framework that replaces discrete latent action encoding with a continuous rotational latent action representation on the group SO(n). This design provides continuity, compositionality, and structured geometry that better capture real‑world action dynamics, and a triplet frame learning scheme enforces meaningful temporal dynamics while preventing degeneration. Trained with 1.7 B parameters on large cross‑embodiment datasets, RotVLA achieves state‑of‑the‑art performance on LIBERO and RoboTwin2.0 benchmarks and shows strong real‑world manipulation results.
arXiv:2609.15189v1 Announce Type: new Abstract: Latent action models (LAMs) learn action representations from unlabeled videos by inferring latent actions from visual transitions and reconstructing f...
arXiv:2607. 28362v1 Announce Type: cross Abstract: We present ShadowDancer, a novel approach to any-action, frame-level control of interactive video world models.