The paper introduces Embodiment‑Classifier Guidance, a method that adapts a frozen flow‑based policy π_0 to new robot embodiments without additional demonstrations or parameter updates. By training a lightweight classifier to distinguish successful action chunks between source and target bodies, the policy’s denoising steps are nudged toward the target distribution using the classifier’s gradient. Experiments on the LIBERO simulation benchmark show that this guidance improves average success rates from 72% to 80% across 12 controller–body pairs, reduces failures by 31%, and demonstrates data‑efficient adaptation with only a few target‑body rollouts.
arXiv:2610.32941v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models perform well on robotic manipulation, yet remain brittle once their own actions carry them into states their trai...
arXiv:2609.26420v1 Announce Type: cross Abstract: Text-to-motion models generate plausible human motion but do not model a robot's dynamics; whole-body tracking controllers execute robot references r...
The paper introduces $ abla$Mimic, a method that uses exact gradients from a differentiable simulator to train humanoid control policies more efficiently. It mitigates gradient explosions by clipping per‑sample gradients and reuses simulated windows via multistep regression, while the NablaEngine GPU simulation engine computes full contact dynamics and adjoints in a single megakernel, achieving an 11.5× speedup over MuJoCo Warp. On the Unitree G1, $ abla$Mimic matches state‑of‑the‑art world‑model performance in 20 minutes and surpasses it in 1.8 hours, outperforming standard first‑order baselines that fail due to exploding gradients.
arXiv:2608. 03483v1 Announce Type: cross Abstract: Existing chunk-based Vision-Language-Action (VLA) models execute a fixed number of actions (i.
arXiv:2607.09042v2 Announce Type: replace Abstract: Reinforcement learning is increasingly used to fine-tune vision-language-action (VLA) models, but robot interaction is expensive and learning becom...
arXiv:2605.18727v2 Announce Type: replace-cross Abstract: Evaluating embodied systems with real dexterous hardware requires more than isolated motor-skill tests: an agent must perceive a changing sce...
Human demonstrations are today used largely for pretraining vision-language-action models, with little ability to use the same data for target-embodiment or task-oriented training, which scalable depl...
The paper investigates a specific failure mode in vision‑language‑action models, termed instruction‑action binding, where models respond to language and vision separately but fail to combine them to select the correct action under counterfactual changes. Through behavioral analyses of fine‑tuned policies, the authors show that failed rollouts often preserve source behavior or switch to other demonstrated tasks, indicating that language is not ignored but mis‑bound. They propose Equivariant Counterfactual Training (ECT), which supplies counterfactual demonstrations and a paired loss to enforce correct action selection, achieving significant performance gains across simulated and real‑world benchmarks.
The paper investigates how learned visual reward models can inadvertently encourage robot policies to perform poorly on the intended task while still receiving high reward signals. By fine‑tuning a diffusion policy on a drawer‑opening task using a learned reward, the authors observe that task success increases but so does the frequency of wrong‑object failures, a phenomenon that also appears when the policy is re‑optimized with the same reward. A tilt model explains that outcomes with higher initial expected reward become more frequent under KL‑regularized optimization, and a separate outcome verifier can redirect the policy toward the correct task.
The paper presents a causal analysis of a compressed VLA policy that performs well in offline tests but fails in closed‑loop execution on a simulated pick‑and‑place task. An 8‑layer distillation of Octo‑Base retains most parameters and passes all offline metrics, yet collapses during deployment, with early stages degrading gradually and final transport failing entirely. The failure is traced to a negative, late‑heavy residual in the action trace, and standard remedies (continued training, offline data, command‑level compensation, clamping) do not restore performance; only a minimal‑pair intervention that mixes deployment‑distribution rollouts with teacher data restores parity with the teacher. whyItMatters":"The study demonstrates that offline validation metrics alone are insufficient to guarantee closed‑loop success for compressed policies, highlighting the need for targeted deployment‑time testing and interventions."
The paper proposes a method to distill world‑model representations into compact Vision‑Language‑Action (VLA) policies. By adding a single feature‑alignment term during VLA training, a frozen world model’s internal features are cached and the student policy learns to match them, eliminating the need for a generative future‑rolling component. The resulting lightweight policy runs in 32 ms on an RTX 5090, achieving high performance on LIBERO and RoboCasa‑GR1, and transfers effectively to real robotic hardware.
SynthDemo‑RL introduces a teacher‑student framework that uses an automated teacher to generate successful manipulation trajectories from simulator‑privileged state, which are then distilled into a Vision‑Language‑Action (VLA) student via supervised fine‑tuning. The student is further refined with PPO using binary task‑success rewards. On the LIBERO‑PRO benchmark, SynthDemo‑RL rescues all 27 previously unsolvable tasks and achieves near‑perfect success rates, matching performance that would otherwise require human demonstrations.