The paper introduces RACE, a rollout‑free metric for selecting robot policy checkpoints that uses held‑out successful trajectories to calibrate action deviations against state‑conditional thresholds. By aggregating these calibrated scores with a running product, RACE reduces the influence of later reference predictions when early deviations occur, and averages the cumulative scores over time and across trajectories to produce a checkpoint score. Experiments in simulation and on real robots show that RACE outperforms baseline methods and its rankings correlate with deployment success across various policy backbones and tasks.
arXiv:2606. 29898v1 Announce Type: cross Abstract: Real-world evaluation is the gold standard for robot policies because it tests them against the physical conditions and deployment challenges they are ultimately designed to handle.
Action-conditioned world models offer a scalable alternative to costly real-world interaction for improving robot policies. However, existing world-model-based methods largely treat world models as fo...
arXiv:2610.20318v1 Announce Type: cross Abstract: Reinforcement-learning policies for robot locomotion can fail at deployment when operating conditions shift (e.g., rough or slippery terrain, payload...
SCAPE is a scenario‑conditioned simulation‑augmented policy evaluation framework that predicts real‑world policy performance for specific scenarios using limited paired simulation‑and‑real samples and extensive simulation rollouts. It corrects sim‑to‑real bias in simulation labels before training the prediction model and calibrates prediction uncertainty via conformal prediction. Experiments on autonomous driving and quadruped velocity tracking show SCAPE reduces scenario‑level prediction error, improves testing sample efficiency, narrows calibrated prediction intervals, and generalizes better to out‑of‑distribution scenarios, enabling fine‑grained deployment strategies.
Prism‑GRPO enhances the GRPO reinforcement‑learning algorithm for vision‑language‑action policies by adding a weighted trajectory‑level execution‑quality score to binary success rewards. This approach splits groups with identical outcomes into a quality spectrum, preserving training signal while ensuring successes always outrank failures. Experiments on four RoboTwin tasks show Prism‑GRPO achieves higher success and quality at matched rollout budgets, reaching target success rates with up to 56% fewer rollouts and mitigating reward‑hacking behaviors that transfer to real‑robot deployment.
arXiv:2608. 19425v1 Announce Type: cross Abstract: Reliable performance evaluation is a central bottleneck for deploying robot-learning policies in real-world conditions.
arXiv:2611.05929v1 Announce Type: cross Abstract: Off-policy deep reinforcement learning trains on random samples from a replay buffer of past experience. In physical robot settings, that buffer is d...
The paper introduces Rollback, a method that uses action‑conditioned world models to backtrack to critical decision points, branch into alternative actions, and compare outcomes to attribute success to individual decisions. By turning world models into counterfactual engines, Rollback provides fine‑grained, action‑level supervision without a separate value model and integrates distributed world‑model search with asynchronous real‑robot execution. Experiments on block extraction, ball tossing, and tabletop curling show that Rollback improves success rates by over 10% compared to the strongest baseline and surpasses human performance on two tasks within eight hours of training.
The paper introduces a policy‑aligned predictive verifier (PAV) that evaluates candidate action prefixes from vision‑language‑action (VLA) policies to determine how far to execute before re‑planning. PAV predicts the consequences of prefixes and checks their compatibility with successful task completion, using trajectory‑level survival supervision and a localization prior to learn local judgments. Experiments on LIBERO, RoboTwin 2.0, and real‑world manipulation demonstrate that PAV improves the trade‑off between success rate and query cost compared to existing adaptive methods.
Reinforcement-learning policies for robot locomotion can fail at deployment when operating conditions shift (e.g., rough or slippery terrain, payload changes, sensing degradation), and retraining or m...
The paper introduces Quality-Controllable Policy (QCP), a robot policy that allows real‑time adjustment of quality attributes such as speed and smoothness during task execution. QCP employs an iterative hindsight relabeling loop, where online rollouts are collected under target quality labels, successful trajectories are added to the dataset, and trajectories are relabeled based on measured quality before fine‑tuning the policy. Experiments on Robomimic, RoboEval, and real‑world manipulation tasks show that QCP improves both task execution quality and success rates, enabling higher throughput in conveyor operations and better stability when carrying fragile objects.
The paper introduces Reliability‑Aware Future Conditioning (RAFC), a method that dynamically estimates how much to trust a generated future video clip and selects the most appropriate temporal hypothesis for robot manipulation tasks. By treating temporal misalignment as a control problem rather than a generation problem, RAFC improves success rates on the CALVIN benchmark and a Franka robot under various timing mismatches, outperforming static and uniform‑averaging baselines. The approach builds on Future‑Experience Conditioning (FEC) and learns solely from task reward without explicit alignment supervision.