# Amazon robotics research

Amazon robotics research covers autonomous mobile robots, control software, computer vision, depth sensing, machine learning, object recognition, and robotic manipulation.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## RePlan-VLA: Progress-Aware Replanning for Long-Horizon Vision-Language-Action Models

DevFeed: [RePlan-VLA: Progress-Aware Replanning for Long-Horizon Vision-Language-Action Models](<https://devfeed.tech/articles/replan-vla-progress-aware-replanning-for-long-horizon-vision-language-action-models-142709.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23843>)

Author: Apoorva Vashisth, Manav Kulshrestha, Damon Conover, Zachary Kingston, Aniket Bera

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Robot Navigation](<https://devfeed.tech/topics/robot-navigation.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [execution](<https://devfeed.tech/tags/execution.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [progress](<https://devfeed.tech/tags/progress.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [robotic-manipulation](<https://devfeed.tech/tags/robotic-manipulation.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>)

### AI overview

RePlan-VLA aims to improve long-horizon robotic manipulation by using a VLA model's short-horizon skills within an execution-aware framework. It monitors robot motion to assess progress and revises the future plan based on observed outcomes, while leaving the underlying VLA policy unchanged and requiring no fine-tuning. The abstract reports success-rate gains over the next strongest baseline of 21.9% on RoboCasa365 and 10.4% on CALVIN, and validation on a Franka Research 3 robot.

### Source excerpt

arXiv:2611.23843v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models have emerged as a promising paradigm for general-purpose robotic manipulation, yet their performance often degrades on long-horizon tasks that require multiple interdependent manipulations. Instead of burdening the VLA policies with long-horizon reasoning, memory, action generation, progress tracking, and error recovery, we argue for leveraging their short-horizon manipulation capabilities within an execution-aware framework. To this end, we introduce \textnormal{\textbf{RePlan-VLA}}, a VLA-agnostic framework that translates a VLA's short-horizon capabilities into reliable long-horizon execution. RePlan-VLA explicitly monitors the robot's motion for progress evaluation, and adaptively revises the future plan based on observed outcomes. We preserve the underlying VLA policy entirely, requiring no additional fine-tuning. Experiments on RoboCasa365 and CALVIN benchmarks show an improvement of 21.9% and 10.4% in success rate over the next strongest baseline. We also validate our approach for real-world deployment on a Franka Research 3 robot.

## Feasibility Ceilings Before Learning: Sizing and Timing Bounds for Drone-Mesh Interception

DevFeed: [Feasibility Ceilings Before Learning: Sizing and Timing Bounds for Drone-Mesh Interception](<https://devfeed.tech/articles/feasibility-ceilings-before-learning-sizing-and-timing-bounds-for-drone-mesh-interception-143333.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.20823>)

Author: Dmytro Babarytskyi

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [World models](<https://devfeed.tech/topics/world-models.md>)

Tags: [bias](<https://devfeed.tech/tags/bias.md>), [control](<https://devfeed.tech/tags/control.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ma](<https://devfeed.tech/tags/cs-ma.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [drone](<https://devfeed.tech/tags/drone.md>), [interception](<https://devfeed.tech/tags/interception.md>), [mesh](<https://devfeed.tech/tags/mesh.md>), [monte-carlo](<https://devfeed.tech/tags/monte-carlo.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [packing](<https://devfeed.tech/tags/packing.md>), [ppo](<https://devfeed.tech/tags/ppo.md>), [probability](<https://devfeed.tech/tags/probability.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [timing](<https://devfeed.tech/tags/timing.md>), [uncertainty](<https://devfeed.tech/tags/uncertainty.md>), [variance](<https://devfeed.tech/tags/variance.md>)

### AI overview

The abstract derives pre-policy bounds for drone-mesh interception and validates them by Monte Carlo. For noisy linear-track extrapolation, least-squares variance propagation plus a delta-method arrival-time term predicts aim-point uncertainty within 1.8% of simulation. A closed-form packed-mesh coverage estimate is accurate to 0.007 absolute. With agents separated by at least 2r, coverage is capped at 90.69% in the plane and 74.05% in a volume; allowing overlap removes this separation-driven ceiling above crossover p=0.80. Among five arrangements, hexagonal packing is best, while centroidal-Voronoi coverage control reaches half its coverage at small budgets. Gaussian uncertainty is conservative for heavy tails and anisotropy but sensitive to bias: a one-sigma aim-point offset costs 14 coverage points. The repositioning commit-time window may be empty. Relative to the applicable ceiling, analytic placement reaches 85% of permitted coverage; centralized PPO averages 15% across three trainer seeds, with paired-bootstrap comparisons against a null policy inconclusive at every seed.

### Source excerpt

arXiv:2610.20823v2 Announce Type: replace-cross Abstract: A mesh of interceptors defending a fixed site must spread over the predicted arrival point of an approaching object. This is a recurring task in multi-agent robotics: N agents, each covering a disc of radius r, spread over a probability density so that a point drawn from it falls within some agent's disc. We give closed forms for the quantities that bound performance before any policy is chosen, and validate each by Monte Carlo. The uncertainty of a point predicted by extrapolating a noisy linear track follows from least-squares variance propagation plus a delta-method term for the uncertain arrival time, agreeing with simulation to within 1.8%. The coverage probability of a packed mesh admits a closed form accurate to 0.007 absolute, with a volumetric analogue. A naive area ratio misses a ceiling: while agents stay at least 2r apart, no mesh exceeds 90.69% coverage in the plane or 74.05% in a volume. This ceiling belongs to the separation constraint, not to the problem - permitting overlap substitutes Kershner's thinnest-covering density and removes it, the better regime above a crossover at p=0.80. Hexagonal packing is optimal among five arrangements; the centroidal-Voronoi configuration coverage control supplies reaches only half its coverage at small budgets. The Gaussian assumption is conservative under heavy tails and anisotropy but fragile under bias: a one-sigma aim-point offset costs 14 points of coverage. A repositioning team must also choose a commit time, and the admissible window can be empty. Against the applicable ceiling, analytic placement attains 85% of what the configuration permits while centralised PPO averages 15% over three trainer seeds, with a paired bootstrap against a null policy containing zero at every seed. We argue such ceilings should be computed before any learning experiment and quoted alongside every result.

## VoxelSBFL: Voxel-Level Spectrum-Based Failure Prediction for Vision-Language-Action Models

DevFeed: [VoxelSBFL: Voxel-Level Spectrum-Based Failure Prediction for Vision-Language-Action Models](<https://devfeed.tech/articles/voxelsbfl-voxel-level-spectrum-based-failure-prediction-for-vision-language-action-models-139607.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22453>)

Author: Arusa Kanwal, Pablo Valle, Paola Roggero, Miren Illarramendi, Shaukat Ali, Aitor Arrieta

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Predictive AI](<https://devfeed.tech/topics/predictive-ai.md>), [reliability-engineering](<https://devfeed.tech/topics/reliability-engineering.md>), [vlm](<https://devfeed.tech/topics/vlm.md>)

Tags: [assessment](<https://devfeed.tech/tags/assessment.md>), [black-box](<https://devfeed.tech/tags/black-box.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [failure](<https://devfeed.tech/tags/failure.md>), [language](<https://devfeed.tech/tags/language.md>), [localization](<https://devfeed.tech/tags/localization.md>), [maps](<https://devfeed.tech/tags/maps.md>), [models](<https://devfeed.tech/tags/models.md>), [performance](<https://devfeed.tech/tags/performance.md>), [policy](<https://devfeed.tech/tags/policy.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [roc](<https://devfeed.tech/tags/roc.md>), [roc-auc](<https://devfeed.tech/tags/roc-auc.md>), [score](<https://devfeed.tech/tags/score.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [spectrum](<https://devfeed.tech/tags/spectrum.md>), [vision](<https://devfeed.tech/tags/vision.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>), [vision-language-action-models](<https://devfeed.tech/tags/vision-language-action-models.md>), [voxel](<https://devfeed.tech/tags/voxel.md>)

### AI overview

VoxelSBFL predicts failure risk in unseen VLA-model executions by mapping recorded tool-center-point trajectories to binary workspace-voxel coverage, then combining coverage with execution outcomes in established spectrum-based suspiciousness formulas. This black-box method needs no internal model access. In simulation, evaluation across four VLA models and 96,000 Pick-Up trajectories found an overall macro ROC-AUC of 0.704; the leading techniques significantly outperformed eight action- and trajectory-based baselines. Performance varied with voxel spatial granularity and training-set size. The results support spatial spectra for execution assessment, while runtime monitoring and policy improvement are proposed as future research.

### Source excerpt

arXiv:2611.22453v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models integrate visual perception, language understanding, and action generation to control robots. Testing these systems produces trajectories and task outcomes, but existing approaches provide limited guidance on using these recorded trajectories to assess subsequent executions or inform policy repair. We conjecture that, for a given policy and task, traversing certain workspace regions is associated with a higher likelihood of failure. To investigate this conjecture, we propose VoxelSBFL, a black-box adaptation of Spectrum-Based Fault Localization (SBFL) in which workspace voxels replace program elements as units of analysis. Recorded tool-center-point (TCP) trajectories provide binary voxel coverage, which is combined with execution outcomes to construct suspiciousness maps using established SBFL formulas. These maps identify failure-associated regions and score unseen trajectories without requiring internal model access. We evaluate 34 spectrum-based techniques across four VLA models using 96,000 Pick-Up trajectories in simulation and compare the leading techniques against eight action- and trajectory-based metrics. VoxelSBFL obtains an overall macro ROC-AUC of 0.704, significantly outperforming state-of-the art baselines. Sensitivity analyses show that performance varies with spatial granularity and training-set size. These findings support spatial spectra for VLA execution assessment and motivate future research on runtime monitoring, self-healing and targeted policy improvement.

## When Backdoors Can Be Switched Off: Suppression-Aware Backdoor Attacks on Vision-Language-Action Models

DevFeed: [When Backdoors Can Be Switched Off: Suppression-Aware Backdoor Attacks on Vision-Language-Action Models](<https://devfeed.tech/articles/when-backdoors-can-be-switched-off-suppression-aware-backdoor-attacks-on-vision-language-action-models-139284.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19774>)

Author: Nur Afsa Syeda, Mohamed Elmahallawy

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [smolvla](<https://devfeed.tech/topics/smolvla.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [attacks](<https://devfeed.tech/tags/attacks.md>), [backdoor](<https://devfeed.tech/tags/backdoor.md>), [conditional](<https://devfeed.tech/tags/conditional.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [detection](<https://devfeed.tech/tags/detection.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [instructions](<https://devfeed.tech/tags/instructions.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [policies](<https://devfeed.tech/tags/policies.md>), [risks](<https://devfeed.tech/tags/risks.md>), [security](<https://devfeed.tech/tags/security.md>), [smolvla](<https://devfeed.tech/tags/smolvla.md>), [trigger](<https://devfeed.tech/tags/trigger.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>)

### AI overview

The study tests VLA backdoors that activate only when multiple conditions are met and can be suppressed by an instruction. Its variants include Cross-Modal Trigger, which combines robot-state and spatially separated visual triggers; Self-Suppressing Trigger, which uses a state trigger plus [DISRUPT] and is suppressed by [RESTORE]; and Multi-Trigger, which adds a visual trigger while allowing suppression even when physical triggers persist. In SmolVLA experiments on two SO-100 datasets, Self-Suppressing and Multi-Trigger reduced Isolation Forest detection from 96-98% to 1-18% and action-space divergence by 10-14x, while reporting 63-71% attack success and perfect clean-task performance. OpenVLA experiments on LIBERO reportedly transferred the attacks to a 7B-parameter policy, evading STRIP and action-distribution-shift detection. The abstract argues conditional suppression should be included when evaluating VLA backdoor threats.

### Source excerpt

arXiv:2611.19774v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models map visual observations, robot states, and natural-language instructions to robotic actions, but their multimodal interfaces introduce new security risks. Existing VLA backdoor attacks primarily study trigger-based activation and stealthy action manipulation, leaving conditional suppression of malicious behavior underexplored. We investigate suppression-aware backdoors, which combine conjunctive trigger activation with instruction-conditioned suppression. We instantiate this design through three variants: Cross-Modal Trigger, combining robot-state and spatially separated visual triggers; Self-Suppressing Trigger, requiring a state trigger and a [DISRUPT] instruction token for activation, with [RESTORE] token suppressing the attack; and Multi-Trigger, additionally requiring a visual trigger while retaining suppression even when physical triggers persist. We evaluate these attacks using SmolVLA on two SO-100 manipulation datasets against four baselines: State Backdoor, BadVLA, TabVLA, and SilentDrift. Self-Suppressing and Multi-Trigger reduce Isolation Forest detection from 96--98% to 1--18% and action-space l divergence by 10--14X, achieving 63--71% attack success rates with perfect clean-task performance. Experiments with OpenVLA on LIBERO further show that the attacks transfer to a 7B-parameter policy, evading STRIP and action-distribution-shift detection. Our findings establish conditional suppression as an important dimension for evaluating backdoor threats in VLA policies.

## E0: Expressive Fine-Grained Discrete Action Prediction for Vision-Language-Action Models via Tweedie Discrete Diffusion

DevFeed: [E0: Expressive Fine-Grained Discrete Action Prediction for Vision-Language-Action Models via Tweedie Discrete Diffusion](<https://devfeed.tech/articles/e0-expressive-fine-grained-discrete-action-prediction-for-vision-language-action-models-via-tweedie-discrete-diffusion-134951.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2511.21542>)

Author: Zhihao Zhan, Jiaying Zhou, Likui Zhang, Qinhan Lyu, Hao Liu, Weizheng Li, Jusheng Zhang, Ziliang Chen, Tianshui Chen, Ruifeng Zhai, Keze Wang, Liang Lin, Guangrun Wang

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [vlm](<https://devfeed.tech/topics/vlm.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [arm](<https://devfeed.tech/tags/arm.md>), [camera](<https://devfeed.tech/tags/camera.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [denoising](<https://devfeed.tech/tags/denoising.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [distributions](<https://devfeed.tech/tags/distributions.md>), [environments](<https://devfeed.tech/tags/environments.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [fine](<https://devfeed.tech/tags/fine.md>), [framework](<https://devfeed.tech/tags/framework.md>), [generation](<https://devfeed.tech/tags/generation.md>), [language](<https://devfeed.tech/tags/language.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [masking](<https://devfeed.tech/tags/masking.md>), [models](<https://devfeed.tech/tags/models.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [vision](<https://devfeed.tech/tags/vision.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>), [vision-language-action-models](<https://devfeed.tech/tags/vision-language-action-models.md>)

### AI overview

E0 generates robotic actions by iteratively denoising quantized action tokens with Tweedie discrete diffusion. The authors motivate this design by action distributions' multiple peaks, token-based reasoning in pretrained VLM/VLA backbones, and the finite resolution of real-world control. They say the method supports fine-grained executable actions while avoiding the distributional mismatch of masking-based discrete diffusion. On LIBERO, VLABench, ManiSkill, and a real-world Franka arm, E0 reportedly achieves state-of-the-art results, outperforming strong baselines by 8.1% on average.

### Source excerpt

arXiv:2511.21542v3 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models offer a unified framework for robotic manipulation by integrating visual perception, language understanding, and control generation. However, existing VLA systems still struggle to generalize across diverse tasks, scenes, and camera viewpoints, and often produce coarse or unstable actions. We argue that these limitations are closely tied to the structural properties of actions in VLA settings, including the inherent multi-peaked nature of action distributions, the token-based symbolic reasoning of pretrained VLM/VLA backbones, and the effective finite resolution imposed by real-world robotic control. Motivated by these properties, we introduce E0, a tweedie discrete diffusion framework that formulates action generation as iterative denoising over quantized action tokens. By operating in a discrete action space with a principled diffusion process, E0 naturally aligns with token-based reasoning, supports fine-grained yet executable action control, and avoids the distributional mismatch of masking-based discrete diffusion. Experiments on LIBERO, VLABench, ManiSkill, and a real-world Franka arm demonstrate that E0 achieves state-of-the-art performance across diverse environments, outperforming strong baselines by 8.1% on average.

## Learning Low-Latency Action Verification for Frozen Vision-Language-Action Models

DevFeed: [Learning Low-Latency Action Verification for Frozen Vision-Language-Action Models](<https://devfeed.tech/articles/learning-low-latency-action-verification-for-frozen-vision-language-action-models-132055.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14834>)

Author: Wen-Han Hsieh, Ken Jiankun Zheng, Chia-Chien Chen, David M. Chan, Trevor Darrell

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [latency](<https://devfeed.tech/tags/latency.md>), [learning](<https://devfeed.tech/tags/learning.md>), [models](<https://devfeed.tech/tags/models.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [robotic-manipulation](<https://devfeed.tech/tags/robotic-manipulation.md>), [verification](<https://devfeed.tech/tags/verification.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>)

### AI overview

HDH uses autonomous rollouts from a frozen vision-language-action policy to train a compact verifier that predicts task success for action chunks. At each policy query, the frozen model generates candidate actions and the verifier selects the highest-scoring chunk for execution, aiming to reduce the latency of computationally expensive verification. The abstract reports an 82.3% average success rate across four LIBERO-10 tasks and a 22.6% average success improvement over the baseline on three real-world manipulation tasks.

### Source excerpt

arXiv:2611.14834v1 Announce Type: cross Abstract: Vision-language-action (VLA) models have made significant progress in general-purpose robotic manipulation. However, sampled action sequences can be ineffective, while evaluating alternatives with computationally expensive verifiers adds latency at each policy query. In this work, we propose HDH, a lightweight framework that uses hindsight from a frozen VLA's own deployments to improve action verification. The framework has three stages. First, we deploy the policy and collect its autonomous rollouts. We then train a compact action verifier to assign each action chunk a predicted task-success score. Finally, the frozen VLA generates action candidates, and the verifier selects the candidate with the highest score for execution. Across four LIBERO-10 tasks, HDH achieves an average success rate of 82.3\% and on three real-world manipulation tasks, HDH improves average success over the baseline policy by 22.6\%.

## Curie's Jarvis: Autonomous Robotic Chemistry Experiments from Printed Protocols

DevFeed: [Curie's Jarvis: Autonomous Robotic Chemistry Experiments from Printed Protocols](<https://devfeed.tech/articles/curie-s-jarvis-autonomous-robotic-chemistry-experiments-from-printed-protocols-128280.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.13251>)

Author: Rumi Loghmani, Bozhang Song, Victor Alvarado, Aliyah Redding, Amir Barati Farimani

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Humanoid Robots](<https://devfeed.tech/topics/humanoid-robots.md>)

Tags: [arm](<https://devfeed.tech/tags/arm.md>), [automation](<https://devfeed.tech/tags/automation.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [chemistry](<https://devfeed.tech/tags/chemistry.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [protocols](<https://devfeed.tech/tags/protocols.md>), [robotic](<https://devfeed.tech/tags/robotic.md>)

### AI overview

Curie's Jarvis lets a Franka Panda robot arm run chemistry protocols from a photo of a printed instruction sheet on an ordinary, unmarked bench. It identifies containers by reading labels from four camera views, relocates targets before each step, and adjusts how actions are performed, including speed, based on what is being handled. In reported trials without human intervention, it completed one children's-kit protocol in 5 of 5 runs and another in 4 of 5; across 160 locates using printed and handwritten labels, it identified every container correctly.

### Source excerpt

arXiv:2611.13251v1 Announce Type: cross Abstract: A person running a bench experiment works from a printed instruction sheet, tells identical containers apart by the labels written on them, and looks at the bench again before each step. Lab automation usually works the other way around: the procedure is written as a script, and containers stand at fixed positions or carry fiducial markers. We present Curie's Jarvis, a system that lets a robot arm carry out a chemistry experiment on a bench set up for people. It reads the steps from a photo of the printed sheet, finds containers on an ordinary bench with no fixed positions or markers, tells identical-looking containers apart by reading their labels in four camera views, locates its target again before every step, and sets how each step is carried out, including its speed, from what is being handled. On a Franka Panda arm, Curie's Jarvis completed 5 of 5 runs of one protocol from a children's chemistry kit and 4 of 5 runs of another, with no human intervention, and in 160 locates on printed and handwritten labels it identified every container correctly. This brings robots a step closer to doing bench chemistry the way people do.

## Robotic Car Charger Locks On With Suction

DevFeed: [Robotic Car Charger Locks On With Suction](<https://devfeed.tech/articles/robotic-car-charger-locks-on-with-suction-124908.md>)

Original publisher: [Read original article](<https://hackaday.com/2026/11/09/robotic-car-charger-locks-on-with-suction/>)

Author: Donald Papp

Published: 2026-11-10T06:00:59Z

Content type: article

Language: en

Sources: [Hackaday](<https://devfeed.tech/sources/hackaday.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [smartdevices](<https://devfeed.tech/topics/smartdevices.md>)

Tags: [car](<https://devfeed.tech/tags/car.md>), [car-charger](<https://devfeed.tech/tags/car-charger.md>), [car-hacks](<https://devfeed.tech/tags/car-hacks.md>), [charger](<https://devfeed.tech/tags/charger.md>), [charuco](<https://devfeed.tech/tags/charuco.md>), [lidar](<https://devfeed.tech/tags/lidar.md>), [machine-vision](<https://devfeed.tech/tags/machine-vision.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [robotic-charger](<https://devfeed.tech/tags/robotic-charger.md>), [robots-hacks](<https://devfeed.tech/tags/robots-hacks.md>), [suction](<https://devfeed.tech/tags/suction.md>), [tesla](<https://devfeed.tech/tags/tesla.md>)

### AI overview

The robot aligns with the car using Mecanum wheels and LiDAR, then fine-tunes its position with a camera and a ChARuco marker beneath the charge port. API calls open the port after the system confirms the car is parked and unused. Because inserting the plug takes more force than the lightweight robot can provide on its own, two suction cups anchor it to the car while a gear assembly pushes the charger in. To park afterward, it reverses the connection process and swivels its camera toward a wall marker, avoiding a full turn. The excerpt describes the robot's design and operation but provides no performance measurements.

### Source excerpt

Not only does [Mew463]'s robotic car charger take care of plugging in his vehicle for him, it has some really clever features that are great to see in action. We've ...read more

## Amazon's Next Home Robot: A Second Chance After Astro's Rocky Start

DevFeed: [Amazon's Next Home Robot: A Second Chance After Astro's Rocky Start](<https://devfeed.tech/articles/amazon-s-next-home-robot-a-second-chance-after-astro-s-rocky-start-124336.md>)

Original publisher: [Read original article](<https://www.techrepublic.com/article/news-amazon-new-home-robot-astro-fauna-robotics/>)

Author: Justin Meyers

Published: 2026-11-09T21:28:31Z

Content type: article

Language: en

Sources: [TechRepublic](<https://devfeed.tech/sources/techrepublic.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Robot Navigation](<https://devfeed.tech/topics/robot-navigation.md>), [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [alexa](<https://devfeed.tech/tags/alexa.md>), [amazon](<https://devfeed.tech/tags/amazon.md>), [amazon-astro](<https://devfeed.tech/tags/amazon-astro.md>), [apple](<https://devfeed.tech/tags/apple.md>), [arm](<https://devfeed.tech/tags/arm.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [astro](<https://devfeed.tech/tags/astro.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [business](<https://devfeed.tech/tags/business.md>), [children](<https://devfeed.tech/tags/children.md>), [cnbc](<https://devfeed.tech/tags/cnbc.md>), [companion](<https://devfeed.tech/tags/companion.md>), [companion-robot](<https://devfeed.tech/tags/companion-robot.md>), [complexity](<https://devfeed.tech/tags/complexity.md>), [consumer-robotics](<https://devfeed.tech/tags/consumer-robotics.md>), [cost](<https://devfeed.tech/tags/cost.md>), [fauna-robotics](<https://devfeed.tech/tags/fauna-robotics.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [home](<https://devfeed.tech/tags/home.md>), [home-robot](<https://devfeed.tech/tags/home-robot.md>), [home-robots](<https://devfeed.tech/tags/home-robots.md>), [humanoid-robots](<https://devfeed.tech/tags/humanoid-robots.md>), [news](<https://devfeed.tech/tags/news.md>), [robot](<https://devfeed.tech/tags/robot.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [robots](<https://devfeed.tech/tags/robots.md>), [smart-home-technology](<https://devfeed.tech/tags/smart-home-technology.md>), [tech-industry](<https://devfeed.tech/tags/tech-industry.md>)

### AI overview

Amazon is building another home robot effort with staff and experience drawn from Astro, but has disclosed no design, capabilities, price, or launch date. Fauna Robotics' acquisition adds its Sprout humanoid and robotics expertise; Fauna says Sprout will remain available to developers and researchers, which does not establish that it will become Amazon's household product. Astro showed both potential uses, such as remote pet and appliance checks, and the limits of an expensive device with narrow functionality and a small invitation-only audience. Domestic robots must handle clutter, people, pets, stairs, and varied layouts; Amazon says a substantial gap remains between controlled labs and everyday spaces. Astro can carry items placed in its bin, but independent retrieval is a different, unannounced capability. The excerpt frames the practical tests for a future product as reliable repeated tasks, accessible pricing and availability, and clear long-term repair, software, and service support.

### Source excerpt

Amazon has another shot at making home robots useful enough to justify their price and wants to apply what it learned from Astro. The post Amazon's Next Home Robot: A Second Chance After Astro's Rocky Start appeared first on TechRepublic.

## HiMEM: Hippocampus-Inspired Memory Consolidation, Pruning, and Composition for Vision-Language-Action Models in 3D Scenes

DevFeed: [HiMEM: Hippocampus-Inspired Memory Consolidation, Pruning, and Composition for Vision-Language-Action Models in 3D Scenes](<https://devfeed.tech/articles/himem-hippocampus-inspired-memory-consolidation-pruning-and-composition-for-vision-language-action-models-in-3d-scenes-124662.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.09355>)

Author: Boran Zhao, Yanbin Hu, Jun Ye, Zhenxian Hu, Yuxin Zhai, Jin Cui, Pengju Ren

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [vlm](<https://devfeed.tech/topics/vlm.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [3D](<https://devfeed.tech/topics/3d.md>), [memory pool](<https://devfeed.tech/topics/memory-pool.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [composition](<https://devfeed.tech/tags/composition.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [language](<https://devfeed.tech/tags/language.md>), [memory](<https://devfeed.tech/tags/memory.md>), [models](<https://devfeed.tech/tags/models.md>), [pruning](<https://devfeed.tech/tags/pruning.md>), [vision](<https://devfeed.tech/tags/vision.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>), [vision-language-action-models](<https://devfeed.tech/tags/vision-language-action-models.md>)

### AI overview

HiMEM targets noisy episodic memory in 3D-scene VLA tasks by consolidating experiences into compact semantic memories, pruning episodes whose content is already captured, and composing multiple episodic and semantic memories for complex scenarios. The authors report improved cross-task and long-horizon performance across VLA backbones; on LIBERO, pi_0's average success rate rises from 91.88% to 98.20%. The supplied text does not give further implementation details or results for LIBERO-Plus.

### Source excerpt

arXiv:2611.09355v1 Announce Type: cross Abstract: Memory plays a crucial role in long-horizon task execution and cross-task generalization for Vision-Language-Action (VLA) models operating in complex 3D scenes. However, existing VLA memory mechanisms suffer from memory noise, particularly in complex 3D scenes, as stored episodic memories contain extensive redundant details about scene layouts, object configurations, and robot-specific states, which obscure generalizable manipulation principles and hinder cross-task transfer. To address this challenge, we first introduce a hippocampus-inspired memory consolidation mechanism that synthesizes extensive episodic experiences into a compact set of essential semantic memories, enabling VLA models to derive generalizable principles from practical experience. Second, we propose an episodic memory pruning mechanism that removes redundant episodic memories whose contents are captured by consolidated semantic memories, thereby increasing effective memory capacity and improving long-horizon task execution. Finally, to handle complex task scenarios, we develop a flexible memory composition mechanism that jointly leverages multiple episodic and semantic memories to enhance cross-scenario generalization. Experiments on LIBERO, LIBERO-Plus demonstrate the effectiveness of HiMEM across different VLA backbones, improving the average success rate of pi_0 from 91.88% to 98.20% on LIBERO.

## IMUDINO: Self-distilled Foundation Model for Inertial Sensing

DevFeed: [IMUDINO: Self-distilled Foundation Model for Inertial Sensing](<https://devfeed.tech/articles/imudino-self-distilled-foundation-model-for-inertial-sensing-120192.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07437>)

Author: Zhoutong Ye, Chengwen Zhang, Zixuan Wang, Xufeng Jian, Zhe He, Qingyang Wan, Xutong Wang, Jiaqi Liu, Mingze Sun, Yu Mei, Chun Yu, Yuanchun Shi

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [features](<https://devfeed.tech/tags/features.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [global](<https://devfeed.tech/tags/global.md>), [health](<https://devfeed.tech/tags/health.md>), [model-architecture](<https://devfeed.tech/tags/model-architecture.md>), [motion](<https://devfeed.tech/tags/motion.md>), [motion-capture](<https://devfeed.tech/tags/motion-capture.md>), [positional-encoding](<https://devfeed.tech/tags/positional-encoding.md>), [recognition](<https://devfeed.tech/tags/recognition.md>), [self-supervised-learning](<https://devfeed.tech/tags/self-supervised-learning.md>), [sensing](<https://devfeed.tech/tags/sensing.md>), [sensor](<https://devfeed.tech/tags/sensor.md>), [ssl](<https://devfeed.tech/tags/ssl.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

IMUDINO adapts DINO self-supervised learning to inertial sensing, using semantics-preserving motion transformations to simulate changes in movement and sensor setup. Its plain transformer encoder uses frequency-aware positional encoding to accept varying input lengths and sampling rates. The abstract reports competitive performance across activity recognition, health, motion capture, and odometry, with features that generalize across users, sensor setups, and tasks; for robot odometry, it also reports generalization across embodiments. The available text is an abstract and does not provide detailed evaluation results.

### Source excerpt

arXiv:2611.07437v1 Announce Type: cross Abstract: We present IMUDINO, the first foundation model (FM) for inertial sensing to achieve competitive performance across diverse sensing tasks including human activity recognition (HAR), health, motion capture, and odometry. Prior inertial FMs often focus on one subfield of inertial sensing and sometimes underperform state-of-the-art supervised methods (e.g., in HAR), making their scope considerably narrower than FMs for vision and natural language. Moreover, they often require fixed input length and sampling rate, and generalize poorly to unseen sensor locations. To build a general-purpose foundation model for inertial sensing, we adapt the DINO self-supervised learning (SSL) paradigm to inertial data through a set of semantics-preserving motion transformations that simulate variations in motion pattern and sensor setup. For model architecture, we use a plain transformer encoder with frequency-aware positional encoding that natively supports varying input length and sampling rate. This simple setup proved remarkably effective for learning global and dense features generalizable across users, sensor setups, tasks, and, in the case of robot odometry, embodiments. We hope IMUDINO can serve as a backbone for inertial sensing and kickstart the development of general-purpose inertial FMs.

## Concertina: Embodying Manipulation and Cognition in One Diffusion Vision-Language Model

DevFeed: [Concertina: Embodying Manipulation and Cognition in One Diffusion Vision-Language Model](<https://devfeed.tech/articles/concertina-embodying-manipulation-and-cognition-in-one-diffusion-vision-language-model-120202.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07513>)

Author: Chengcheng Wang, Wei He, Wenhao Li, Hongguang Li, Kai Han, Jianyuan Guo, Chang Xu

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [vlm](<https://devfeed.tech/topics/vlm.md>), [diffusion-transformers](<https://devfeed.tech/topics/diffusion-transformers.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [blocks](<https://devfeed.tech/tags/blocks.md>), [concertina](<https://devfeed.tech/tags/concertina.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [control](<https://devfeed.tech/tags/control.md>), [coupling](<https://devfeed.tech/tags/coupling.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [generation](<https://devfeed.tech/tags/generation.md>), [generator](<https://devfeed.tech/tags/generator.md>), [interface](<https://devfeed.tech/tags/interface.md>), [interfaces](<https://devfeed.tech/tags/interfaces.md>), [layers](<https://devfeed.tech/tags/layers.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [model](<https://devfeed.tech/tags/model.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [operator](<https://devfeed.tech/tags/operator.md>), [vision](<https://devfeed.tech/tags/vision.md>), [vision-language-model](<https://devfeed.tech/tags/vision-language-model.md>)

### AI overview

Concertina embeds manipulation in a shared diffusion vision-language model: each control timestep becomes a continuous latent action token, and the shared Transformer reconstructs ordered action blocks using the corrupt-and-reconstruct operator from multimodal pretraining. The design applies physical-coordinate supervision on the shared model's computation and parameter path, with no separate action-generator layers; active action projections account for 0.21% of its 4.46B parameters, compared with 16.5%-35.1% dedicated action-expert capacity in representative continuous VLAs. An Action Interface Audit finds cross-time trajectory coupling in the backbone's middle and late layers. The abstract reports 98.15% average success on regular LIBERO, 85.9% on LIBERO-Plus, and 4.142 average completed chain length on CALVIN ABC-D; it provides no experimental details beyond these summary results.

### Source excerpt

arXiv:2611.07513v1 Announce Type: cross Abstract: Most vision-language-action (VLA) policies decouple manipulation from cognition by bolting an external action mechanism onto a vision-language model (VLM)--either quantizing actions into discrete tokens whose errors bear no stable correspondence to physical displacements, or delegating continuous trajectory generation to a separate action expert whose gradients compete with or are blocked from updating the VLM backbone. We argue that manipulation and cognition should instead be embodied in the same model, shifting from a decoupled VLA to an Embodied Vision-Language Model (eVLM). We introduce Concertina, an eVLM that places continuous trajectory representation, blockwise generation, and physical-coordinate supervision directly on the computation and parameter path of a shared diffusion VLM backbone. Each control timestep is encoded as one continuous latent action token in the backbone hidden space, and the shared Transformer reconstructs temporally ordered action blocks via the same corrupt-and-reconstruct operator used for multimodal pretraining. Active action projections account for only 0.21% of parameters in the 4.46B model with zero separate action-generator layers, compared with 16.5%-35.1% dedicated action-expert capacity in representative continuous VLAs. Using our proposed Action Interface Audit, we confirm that cross-time trajectory coupling emerges in the middle and late layers of the shared backbone rather than at the action interfaces. On standard benchmarks, Concertina achieves 98.15% average success on regular LIBERO, 85.9% on the 10,030-task LIBERO-Plus benchmark, and 4.142 average completed chain length on CALVIN ABC-D. Code: https://github.com/lose4578/Concertina Hugging Face: https://huggingface.co/collections/Traly/concertina

## Dexterous Robotic Hands in the Foundation-Model Era: A Full-Stack Survey of Hardware, Data, Cross-Embodiment Transfer, Learning, and Evaluation

DevFeed: [Dexterous Robotic Hands in the Foundation-Model Era: A Full-Stack Survey of Hardware, Data, Cross-Embodiment Transfer, Learning, and Evaluation](<https://devfeed.tech/articles/dexterous-robotic-hands-in-the-foundation-model-era-a-full-stack-survey-of-hardware-data-cross-embodiment-transfer-learning-and-evaluation-118501.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.06724>)

Author: Ji Pei

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Dataset development](<https://devfeed.tech/topics/dataset-development.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [data](<https://devfeed.tech/tags/data.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [full-stack](<https://devfeed.tech/tags/full-stack.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [human-data](<https://devfeed.tech/tags/human-data.md>), [interfaces](<https://devfeed.tech/tags/interfaces.md>), [learning](<https://devfeed.tech/tags/learning.md>), [november-2026](<https://devfeed.tech/tags/november-2026.md>), [open-problems](<https://devfeed.tech/tags/open-problems.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [sensing](<https://devfeed.tech/tags/sensing.md>), [survey](<https://devfeed.tech/tags/survey.md>), [synthesis](<https://devfeed.tech/tags/synthesis.md>), [transfer-learning](<https://devfeed.tech/tags/transfer-learning.md>)

### AI overview

The survey's central finding is that dexterous-hand progress depends on choices across the full system: human data is useful when its kinematics, visuals and contact align with the robot; shared representations and foundation policies still face hand-specific limits in force, latency and contact; and task completion alone does not show physical correctness or deployment readiness. It reviews hardware, data, cross-embodiment transfer, learning, evaluation and applications, and distinguishes experimental comparisons from corpus statistics, vendor specifications and company-reported results. Evidence remains limited for zero-shot transfer of contact-rich skills across different hand morphologies, long-duration reliability and cross-sensor tactile generalization. This is a structured narrative review, not an exhaustive systematic review or pooled performance ranking.

### Source excerpt

arXiv:2611.06724v1 Announce Type: cross Abstract: Dexterous robotic hands are increasingly coupled to large-scale human data and foundation models, yet progress in hardware, representation learning and evaluation is often assessed separately. This survey reviews work from January 2024 to early November 2026, with earlier studies included to establish technical lineages. We organize the literature into five interdependent layers: hardware; data collection and curation; cross-embodiment transfer; learning and foundation models; and evaluation and applications. Human hand-object understanding and tactile sensing provide two cross-cutting perspectives. A system-level architecture, comparative tables and five case analyses connect design choices across these layers. We distinguish within-study experimental comparisons from heterogeneous corpus statistics, vendor specifications and company-reported results. Three conclusions emerge. First, the value of human data depends on kinematic, visual and contact alignment, not recording volume alone. Second, shared action representations and foundation policies do not remove embodiment-specific constraints on force, latency and contact. Third, task completion is insufficient to establish physical correctness or deployment readiness. Evidence remains limited for zero-shot transfer of contact-rich skills between morphologically different hands, long-duration reliability, and cross-sensor tactile generalization. We identify eleven open problems and propose reporting interfaces and controlled comparisons that can make full-stack progress more measurable. The review is a structured narrative synthesis rather than a pooled performance ranking or an exhaustive systematic review.

## Leap-VLA: High-Rate Robot Control from a Low-Rate VLM Backbone

DevFeed: [Leap-VLA: High-Rate Robot Control from a Low-Rate VLM Backbone](<https://devfeed.tech/articles/leap-vla-high-rate-robot-control-from-a-low-rate-vlm-backbone-115118.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01011>)

Author: Lin Yang, Yunjie Shen, Zheyu Li, Zhenjie Tian, Yongli Wang, Dongze Huang, Weiqi Wang, Yucheng Zhu, Zhaoquan Jiang, Guofeng Xie, Peigen Fang

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [vlm](<https://devfeed.tech/topics/vlm.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [post-training](<https://devfeed.tech/topics/post-training.md>), [Cache](<https://devfeed.tech/topics/cache.md>), [Image processing](<https://devfeed.tech/topics/image-processing.md>)

Tags: [1-6](<https://devfeed.tech/tags/1-6.md>), [2](<https://devfeed.tech/tags/2.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [adaptive](<https://devfeed.tech/tags/adaptive.md>), [added](<https://devfeed.tech/tags/added.md>), [anchor](<https://devfeed.tech/tags/anchor.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [asynchronous](<https://devfeed.tech/tags/asynchronous.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cached-context](<https://devfeed.tech/tags/cached-context.md>), [chain-of-thought](<https://devfeed.tech/tags/chain-of-thought.md>), [content](<https://devfeed.tech/tags/content.md>), [context](<https://devfeed.tech/tags/context.md>), [control](<https://devfeed.tech/tags/control.md>), [cost](<https://devfeed.tech/tags/cost.md>), [covers](<https://devfeed.tech/tags/covers.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [robot](<https://devfeed.tech/tags/robot.md>), [vla](<https://devfeed.tech/tags/vla.md>), [vlm](<https://devfeed.tech/tags/vlm.md>)

### AI overview

Leap-VLA separates slow semantic understanding from fast robot control: a VLM refreshes cached context infrequently, while a flow-matching action expert combines it with fresh images from a compact robot-video encoder. The expert receives cache age as an input, and training uses real cache-age histories plus increasing pixel corruption to account for stale context. On an asynchronous deployment stack, success averaged 95.2% across four LIBERO suites with 1-second semantic refreshes and 94.5% with 2-second refreshes, versus a 96.3% frequency-matched anchor. A forward-only "soft-nudge" enables per-step replanning at zero added latency, with 97.1% success. A chain-of-thought verdict gate also freezes execution on instruction-scene mismatches without reducing success on feasible tasks. These results are from the reported LIBERO evaluations; the supplied text does not establish performance beyond them.

### Source excerpt

arXiv:2611.01011v1 Announce Type: new Abstract: Vision-language-action (VLA) models re-encode the full observation through a multi-billion-parameter vision-language model (VLM) at every control step, even though scene semantics evolve at roughly 1 Hz while reactive control requires 20-50 Hz. We present a VLA that decouples the two: a VLM runs as a low-frequency semantic anchor whose cached context is reused across control steps, while a flow-matching action expert operates at high frequency, consuming the cached semantics with a fresh, low-latency visual stream from a compact encoder pretrained on robot video with masked reconstruction, optical flow, and depth. The measured age $\tau$ of the cache conditions the expert through adaptive layer normalization. Reusing the cache makes the policy a function of its age, and training covers that function explicitly: cache ages sampled from real episode history, plus an age-ramped pixel corruption that erases the stale snapshot the way deployed staleness corrupts content. On the asynchronous deployment stack, the recipe closes most of the staleness gap: on LIBERO-Spatial, 61.4% to 95.5% at a 2 s semantic refresh (95.9% at 1 s, within 1.6 points of the frequency-matched anchor); averaged over the four LIBERO suites, the deployed policy holds 95.2% at a 1 s refresh and 94.5% at 2 s against a 96.3% matched anchor. A forward-only chunk-boundary correction (soft-nudge), an autograd-free reformulation of real-time-chunking guidance, carries per-control-step re-planning without loss (97.1%) at zero added latency. The same lens exposes a second defect: under an instruction-scene mismatch the policy keeps manipulating the scene, and a chain-of-thought verdict gate trained on zero-cost negative episodes freezes execution exactly on mismatched instructions without hurting feasible success.

## RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory

DevFeed: [RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory](<https://devfeed.tech/articles/roboarchitect-automating-the-design-of-robotic-agents-with-persistent-memory-112147.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.42141>)

Author: Shengguang Wu, Suning Huang, Cherie Ho, Ria Doshi, Hao Zhu, Serena Yeung-Levy

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Robot Navigation](<https://devfeed.tech/topics/robot-navigation.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [recording](<https://devfeed.tech/topics/recording.md>)

Tags: [agent-harness](<https://devfeed.tech/tags/agent-harness.md>), [agents](<https://devfeed.tech/tags/agents.md>), [automated](<https://devfeed.tech/tags/automated.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [design](<https://devfeed.tech/tags/design.md>), [memory](<https://devfeed.tech/tags/memory.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [persistent-memory](<https://devfeed.tech/tags/persistent-memory.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [robotic-manipulation](<https://devfeed.tech/tags/robotic-manipulation.md>)

### AI overview

RoboArchitect improves robot manipulation success by iteratively revising the control harness around a frozen policy, without changing model weights. A coding model inspects rollouts, diagnoses failures, and optimizes the harness on selected episodes; performance is then evaluated on held-out scenes and tasks. Across multiple benchmarks and policies, reported average success rises from 9.7% to 43.3% on unseen scenes and from 12.7% to 40.2% on unseen tasks. The harness can devise control strategies such as clearing a jammed arm or approaching horizontally around an obstacle, and retain experience such as effective grasp depths in persistent memory. Gains are largest when the benchmark differs substantially from the policy's training data; in those cases the harness-wrapped general-purpose policy often outperforms its finetuned version.

### Source excerpt

arXiv:2610.42141v1 Announce Type: cross Abstract: Recent work shows that a frozen robot policy performs far better inside an agent harness, where control code decides when to query the policy, consult vision tools, or command the arm directly. Yet today these harnesses are built by hand and redesigned for each new robot or benchmark. We show that harness design can be automated, offering a strong alternative to finetuning on a novel setup. In RoboArchitect, a coding model observes a robot's rollouts, diagnoses failures, and iteratively revises the harness around the frozen policy. The harness evolves on a set of episodes for its optimization and is evaluated on held-out scenes and tasks. Without updating any model weights, RoboArchitect raises average success from 9.7% to 43.3% on unseen scenes (4.5x) and from 12.7% to 40.2% on unseen tasks (3.2x), across multiple benchmarks and policies. The gains are largest where the benchmark is far from the policy's training data: wrapping a general-purpose policy, the evolved harness yields a specialist that often outperforms even the finetuned version. Intuitive control strategies are automatically designed, such as lifting a jammed arm clear or reaching in horizontally when an obstacle blocks the way from above. The harness also evolves its own persistent memory, carrying learned experience, such as effective grasp depths, into later decisions. These results suggest that substantial headroom in robotic manipulation lies in the agent harness, and its design can be effectively automated and optimized without weight updates.

## I3L: Iterative Interactive Imitation Learning for Whole-Body Mobile Manipulation

DevFeed: [I3L: Iterative Interactive Imitation Learning for Whole-Body Mobile Manipulation](<https://devfeed.tech/articles/i3l-iterative-interactive-imitation-learning-for-whole-body-mobile-manipulation-111935.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.17748>)

Author: Wensi Ai, Christopher Luey, Karthik Dharmarajan, Evans Han, Yunfan Jiang, Stef Ren, Jianwen Xie, Jiajun Wu, Li Fei-Fei, Ruohan Zhang

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [amplify](<https://devfeed.tech/tags/amplify.md>), [arm](<https://devfeed.tech/tags/arm.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [base](<https://devfeed.tech/tags/base.md>), [body](<https://devfeed.tech/tags/body.md>), [can](<https://devfeed.tech/tags/can.md>), [combines](<https://devfeed.tech/tags/combines.md>), [cost](<https://devfeed.tech/tags/cost.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [dagger](<https://devfeed.tech/tags/dagger.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [imitation](<https://devfeed.tech/tags/imitation.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [learning](<https://devfeed.tech/tags/learning.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [mobile](<https://devfeed.tech/tags/mobile.md>), [mobile-manipulation](<https://devfeed.tech/tags/mobile-manipulation.md>)

### AI overview

I3L combines a low-cost, open-source JoyLo+ interface for coordinated whole-body corrections with iterative deployment and retraining, letting operators target failures of the updated policy. In evaluations across three robot embodiments and eight simulated and real-world tasks, four-round HG-DAgger achieved the highest final success in controlled simulations on both tested tasks. Under this protocol, autonomous success rose by an average of 68.3 percentage points over base policies and reached at least 90% on every task; it exceeded demonstration-only training by 35.6 points on average with the same number of training trajectories. Average takeover duration per episode fell 72% over four rounds. The abstract does not provide further experimental details.

### Source excerpt

arXiv:2610.17748v2 Announce Type: replace-cross Abstract: Policies trained offline from teleoperated demonstrations can degrade outside their training distribution. This gap is particularly acute in long-horizon whole-body mobile manipulation, where long horizons amplify compounding errors and coordinated arm, torso, and base motion makes exhaustive demonstration coverage impractical. We introduce I3L, an integrated hardware and learning system that combines whole-body correction with iterative policy refinement to address this last-mile reliability gap. Its low-cost, open-source JoyLo+ interface uses policy-tracking leader arms and touch-triggered takeover to enable coordinated corrections across the body parts, while iterative deployment and retraining allow operators to target failures of the updated policy. We evaluate I3L across three robot embodiments and eight simulated and real-world tasks, including four long-horizon whole-body mobile manipulation tasks. Controlled simulation experiments show that four-round HG-DAgger achieves the highest final success on both tasks. Under this protocol, I3L improves autonomous success over the base policies by an average of 68.3 percentage points and achieves at least 90% success on every task, outperforming demonstration-only training by 35.6 percentage points on average with the same number of training trajectories. Averaged across tasks, takeover duration per episode decreases by 72% over four rounds. Together, these results show that I3L helps close the last-mile gap between offline training and reliable deployment and provides practical insights into collecting and using corrections for long-horizon whole-body mobile manipulation. Videos, visualizations, and appendix are at https://i-3l.github.io/.

## Pushing Dexterity and Safety for In-Hand Object Pose Reaching That Works Remarkably Well

DevFeed: [Pushing Dexterity and Safety for In-Hand Object Pose Reaching That Works Remarkably Well](<https://devfeed.tech/articles/pushing-dexterity-and-safety-for-in-hand-object-pose-reaching-that-works-remarkably-well-104112.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37652>)

Author: Ian Chuang, Hiya Shah, Vamsi Chilamkur, Iman Soltani, Claire Tomlin

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [based](<https://devfeed.tech/tags/based.md>), [camera](<https://devfeed.tech/tags/camera.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [cs-sy](<https://devfeed.tech/tags/cs-sy.md>), [dexterity](<https://devfeed.tech/tags/dexterity.md>), [eess-sy](<https://devfeed.tech/tags/eess-sy.md>), [filter](<https://devfeed.tech/tags/filter.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [motion-capture](<https://devfeed.tech/tags/motion-capture.md>), [object](<https://devfeed.tech/tags/object.md>), [performance](<https://devfeed.tech/tags/performance.md>), [platform](<https://devfeed.tech/tags/platform.md>), [portable](<https://devfeed.tech/tags/portable.md>), [reduce](<https://devfeed.tech/tags/reduce.md>), [release](<https://devfeed.tech/tags/release.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [safety](<https://devfeed.tech/tags/safety.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [spin](<https://devfeed.tech/tags/spin.md>), [study](<https://devfeed.tech/tags/study.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

The study trains a LEAP hand in simulation, without human data, to reach long sequences of object poses across wrist orientations. Diverse resets and goals, generated grasp datasets, and an adaptive friction curriculum support training; careful system identification enables zero-shot transfer to the real hand using either motion capture or a wrist-mounted depth camera. In 13.3 hours across 4,800 unseen goals, four objects, and varied wrist orientations, the state-based policy reached 95.2% of goals within 2 cm and 0.4 rad. A learned residual safety filter reduced object drops by 75% while retaining a 91.5% goal success rate, allowing adjustable performance-safety tradeoffs. The abstract also reports robustness to perturbations and fast arm motion, and demonstrations such as spinning a lightbulb and screwing one in.

### Source excerpt

arXiv:2610.37652v1 Announce Type: cross Abstract: Dexterous manipulation with multi-fingered hands is the next frontier of robotics, yet learned hand skills remain limited to simple grasps or basic repeated motions, such as rotating an object. As robots attempt more demanding in-hand maneuvers, safety, such as not dropping the object, becomes critical, yet it is rarely measured. We study in-hand object pose reaching, a task that demands diverse and difficult grasps: moving an object to almost any position and orientation within reach of the fingers, at any wrist orientation, through long sequences of consecutive goals. We present a recipe for training such policies for a LEAP hand in simulation without human data, built on a range of insights, from diverse resets and goals via generated grasp datasets to an adaptive friction curriculum. With careful system identification, the policies transfer zero-shot to the real hand, using either motion capture or a single wrist-mounted depth camera, and prove remarkably capable and robust. On top of the task policy, we add a learned residual safety filter with adjustable conservativeness, letting us tune the balance between performance and safety and greatly reduce object drops. In 13.3 hours of real-robot evaluation over 4,800 unseen goals across 4 objects and diverse wrist orientations, the state-based policy reaches 95.2% of goals within 2 cm and 0.4 rad, and the filter cuts its drops by 75% while still reaching 91.5% of goals. The policies can also spin a lightbulb held in an arbitrary pose, reach out-of-distribution goals, withstand fast arm motion and perturbations, and, needing only a wrist camera, serve as portable skills such as screwing in a lightbulb. We release the full recipe as a platform for studying performance, reliability, and safety in high-dimensional dexterous systems. Videos and more are available at https://ian-chuang.github.io/dexsafe/.

## DRIFT: A Residual Flow Adapter for Decoding Continuous Outputs in Vision-Language Models

DevFeed: [DRIFT: A Residual Flow Adapter for Decoding Continuous Outputs in Vision-Language Models](<https://devfeed.tech/articles/drift-a-residual-flow-adapter-for-decoding-continuous-outputs-in-vision-language-models-102975.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.05758>)

Author: Zhuoming Liu, Jinhong Lin, Kwan Man Cheng, Lin Zhang, Shayok Bagchi, Yin Li

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [vlm](<https://devfeed.tech/topics/vlm.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [adapter](<https://devfeed.tech/tags/adapter.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generative](<https://devfeed.tech/tags/generative.md>), [grounding](<https://devfeed.tech/tags/grounding.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [planning](<https://devfeed.tech/tags/planning.md>), [pretraining](<https://devfeed.tech/tags/pretraining.md>), [regression](<https://devfeed.tech/tags/regression.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [temporal](<https://devfeed.tech/tags/temporal.md>), [text](<https://devfeed.tech/tags/text.md>), [tokens](<https://devfeed.tech/tags/tokens.md>), [vision-language-models](<https://devfeed.tech/tags/vision-language-models.md>), [vision-language-models-vlms](<https://devfeed.tech/tags/vision-language-models-vlms.md>), [vlms](<https://devfeed.tech/tags/vlms.md>)

### AI overview

DRIFT adapts pretrained vision-language models for precise continuous outputs by combining a base predictor's coarse estimate with a flow-matching module that iteratively refines the residual. This shifts generative modeling from the full output distribution to residuals around a prior, which the authors say simplifies optimization. They report that DRIFT outperforms regression and generative baselines across visual grounding and robotic control tasks and architectures including MLLMs, VLAs, and WAMs. The supplied text does not provide quantitative results or implementation details.

### Source excerpt

arXiv:2606.05758v2 Announce Type: replace-cross Abstract: Many modern vision-language models (VLMs) build on autoregressive decoding of discrete tokens. While text-based output interfaces enable scalable pretraining and strong zero-shot generalization across diverse tasks, they are poorly suited for problems that require precise continuous outputs, such as localizing temporal boundaries of events or generating robotic control actions. To address this challenge, we propose DRIFT, a general framework for adapting pretrained VLMs to continuous decoding tasks. DRIFT combines a base predictor, which provides a coarse estimate of the target output, with a generative refinement module based on flow matching that iteratively improves the prediction. This residual formulation transforms the generative modeling problem from learning a global output distribution to modeling a localized residual distribution around a strong prior, substantially simplifying optimization. We evaluate DRIFT on both perception and planning tasks, including visual grounding and robotic control. Across multiple tasks and architectures spanning MLLMs, VLAs, and WAMs, DRIFT consistently outperforms a strong set of regression- and generative-based solutions.

## Kinematic Constraint-Referenced Evaluation for Articulated Object Manipulation

DevFeed: [Kinematic Constraint-Referenced Evaluation for Articulated Object Manipulation](<https://devfeed.tech/articles/kinematic-constraint-referenced-evaluation-for-articulated-object-manipulation-103287.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37550>)

Author: Baptiste Bonin, Caro Strickland, Audrey Durand

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [constraint](<https://devfeed.tech/tags/constraint.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>)

### AI overview

Constraint-Referenced Strategy (CRS) evaluates whether an articulated-object manipulation policy moves in a way the object's joint permits. It projects the commanded end-effector twist onto the joint axis to separate progress along the joint from off-constraint motion. CRS requires a joint model and a scripted controller derived from it, but no recorded demonstrations or learned judge. Across 74 simulated objects, binary success reflected three transfer failures only as lower success rates; CRS's two channels gave each a distinct signature. A door-trained behavior-cloning policy opened drawers in 59 of 110 episodes, with about 40 times the off-axis rotation of a drawer-trained policy on successful passes (1.27 vs. 0.032 rad/s). Against a scripted controller scored using the same axis, CRS verdicts were unchanged with joint-axis errors up to 30°; absolute thresholds tolerated about 8°.

### Source excerpt

arXiv:2610.37550v1 Announce Type: cross Abstract: Manipulation policies are evaluated predominantly by binary task success, which cannot tell whether the commanded gesture is the one the object's kinematics admit. We introduce Constraint-Referenced Strategy (CRS), which projects the commanded end-effector twist onto the object's joint axis, separating progress along the joint from off-constraint waste. CRS requires a joint model and a joint-following scripted controller derived from it, but no recorded demonstration or learned judge. On 74 simulated articulated objects, success registers three transfer failures as lower rates but attributes none, while the two CRS channels give each a distinct signature. A door-trained behavior-cloning policy opens drawers in 59 of 110 episodes, and in these passes it commands about $40\times$ the off-axis rotation of a drawer-trained policy's passes ($1.27$ vs.\ $0.032$~rad/s). Relative to a scripted controller scored with the same axis, CRS verdicts are unchanged under joint-axis errors up to $30^\circ$; absolute thresholds tolerate about $8^\circ$.

## If at First You Don't Succeed, Backtrack: Lightweight Recovery for Frozen VLAs

DevFeed: [If at First You Don't Succeed, Backtrack: Lightweight Recovery for Frozen VLAs](<https://devfeed.tech/articles/if-at-first-you-don-t-succeed-backtrack-lightweight-recovery-for-frozen-vlas-99680.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.32941>)

Author: Thomas Landeg, Thierry Tambe

Published: 2026-10-26T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [smolvla](<https://devfeed.tech/topics/smolvla.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [autoregressive](<https://devfeed.tech/tags/autoregressive.md>), [control](<https://devfeed.tech/tags/control.md>), [controller](<https://devfeed.tech/tags/controller.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [detection](<https://devfeed.tech/tags/detection.md>), [experts](<https://devfeed.tech/tags/experts.md>), [failure](<https://devfeed.tech/tags/failure.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [lightweight](<https://devfeed.tech/tags/lightweight.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [model](<https://devfeed.tech/tags/model.md>), [policies](<https://devfeed.tech/tags/policies.md>), [policy](<https://devfeed.tech/tags/policy.md>), [recovery](<https://devfeed.tech/tags/recovery.md>), [retry](<https://devfeed.tech/tags/retry.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [robotic-manipulation](<https://devfeed.tech/tags/robotic-manipulation.md>), [training](<https://devfeed.tech/tags/training.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>), [vla](<https://devfeed.tech/tags/vla.md>)

### AI overview

Backtrack adds a small failure-prediction head to frozen vision-language-action policies and uses it to trigger recovery: when predicted failure exceeds a threshold, a controller pauses the policy, retraces recent arm motion, then lets the unchanged policy retry from an earlier configuration. The head uses existing model representations, is trained on outcome-labeled rollouts, and adds 0.48-0.88% to policy parameters. On LIBERO-Long, it reduced failure rates by 19-48% across three tested VLAs, raising success by up to 7.4 percentage points; effects were smaller on shorter suites with 85-99% baseline success. All 12 tested SmolVLA controller configurations improved on the unassisted policy on MetaWorld.

### Source excerpt

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 training data barely covers. Recovering from such failures usually requires recovery demonstrations, policy retraining, or a separate model at inference time. We introduce Backtrack, a lightweight failure detection and recovery mechanism for frozen VLAs that needs none of these. A small failure prediction head, trained only on outcome-labeled rollouts of the frozen policy, reads the VLA's existing internal representations to estimate whether the current episode will fail. When the predicted failure probability exceeds a threshold, a controller pauses the policy, retraces the arm's recently achieved motion, and hands control back to the unchanged policy so it can retry from an earlier configuration. The head adds only 0.48-0.88% to the policy's parameter count. We evaluate Backtrack on LIBERO with three VLAs spanning two action-generation families: SmolVLA and $\pi$0.5, which use flow-matching action experts, and the autoregressive $\pi$0-FAST. On LIBERO-Long, the suite of long-horizon multi-step tasks and the one on which all three policies fail most often, Backtrack reduces each policy's failure rate by 19-48%, raising success by up to 7.4 percentage points; on the three shorter suites, where control success is already 85-99%, its effect is smaller. On MetaWorld, all twelve controller configurations tested with SmolVLA improve on the unassisted policy.

## FocusVLA: Future-Intent Guided Dynamic View Routing for Vision-Language-Action Models

DevFeed: [FocusVLA: Future-Intent Guided Dynamic View Routing for Vision-Language-Action Models](<https://devfeed.tech/articles/focusvla-future-intent-guided-dynamic-view-routing-for-vision-language-action-models-98216.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.31423>)

Author: Huixin Xu, Ziya Yu, Peng Li

Published: 2026-10-23T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [vlm](<https://devfeed.tech/topics/vlm.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>), [Predictive AI](<https://devfeed.tech/topics/predictive-ai.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [architectures](<https://devfeed.tech/tags/architectures.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [exploit](<https://devfeed.tech/tags/exploit.md>), [framework](<https://devfeed.tech/tags/framework.md>), [frameworks](<https://devfeed.tech/tags/frameworks.md>), [future](<https://devfeed.tech/tags/future.md>), [generation](<https://devfeed.tech/tags/generation.md>), [global](<https://devfeed.tech/tags/global.md>), [guide](<https://devfeed.tech/tags/guide.md>), [information](<https://devfeed.tech/tags/information.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [language](<https://devfeed.tech/tags/language.md>), [models](<https://devfeed.tech/tags/models.md>), [routing](<https://devfeed.tech/tags/routing.md>), [view](<https://devfeed.tech/tags/view.md>), [vision](<https://devfeed.tech/tags/vision.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>), [vision-language-action-models](<https://devfeed.tech/tags/vision-language-action-models.md>)

### AI overview

FocusVLA addresses VLA policies' reported tendency to rely persistently on wrist-camera views by routing views according to manipulation stage. It predicts future interaction intent from historical actions, then uses that prediction during action generation to coordinate global and local observations. The abstract reports a 10.10 percentage-point average success-rate improvement over baseline π0.5 on RoboTwin 2.0 and a real-world increase from 71.7% to 85.0%. It also says the intent-prediction and routing modules can be integrated into existing VLA frameworks without changing their core architectures.

### Source excerpt

arXiv:2610.31423v1 Announce Type: cross Abstract: Vision-language-action (VLA) models have become an important paradigm for robotic manipulation by unifying visual perception, language understanding, and action generation. Since reliable manipulation requires sufficient observation of objects, end-effectors, and their spatial relationships, multi-view inputs are widely used to provide complementary spatial information. However, our experiments show that existing VLA models do not fully exploit such multi-view complementarity. The policy tends to persistently rely on the wrist view, lacking the ability to dynamically adjust global-local view dependency across manipulation stages, which limits stage adaptability and robustness. To address this issue, we propose FocusVLA, a stage-aware multi-view routing framework. FocusVLA predicts future interaction intent from historical actions and uses it to guide dynamic view routing during action generation. By adaptively coordinating global and local observations across different task stages, FocusVLA improves multi-view utilization. On RoboTwin 2.0, FocusVLA improves the average success rate over the strongest baseline $\pi_{0.5}$ by 10.10 pp. In real-world evaluation, FocusVLA improves the success rate from 71.7\% to 85.0\%, achieving a 13.3\% gain over $\pi_{0.5}$. Moreover, the intent prediction and dynamic routing modules of FocusVLA can be integrated into existing VLA frameworks without modifying their core architectures.

## Post-hoc Epistemic Uncertainty for Vision-Language-Action Policies

DevFeed: [Post-hoc Epistemic Uncertainty for Vision-Language-Action Policies](<https://devfeed.tech/articles/post-hoc-epistemic-uncertainty-for-vision-language-action-policies-94981.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.29730>)

Author: Johannes Keller, Albrecht M. Blo{\\ss}e, Adrian Lindenmeyer, Thomas Neumuth, Daniel Schneider

Published: 2026-10-22T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [smolvla](<https://devfeed.tech/topics/smolvla.md>), [Amazon robotics research](<https://devfeed.tech/topics/amazon-robotics-research.md>)

Tags: [acquisition](<https://devfeed.tech/tags/acquisition.md>), [adaptive](<https://devfeed.tech/tags/adaptive.md>), [camera](<https://devfeed.tech/tags/camera.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [decoder](<https://devfeed.tech/tags/decoder.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [policies](<https://devfeed.tech/tags/policies.md>), [recognition](<https://devfeed.tech/tags/recognition.md>), [robot](<https://devfeed.tech/tags/robot.md>), [setup](<https://devfeed.tech/tags/setup.md>), [smolvla](<https://devfeed.tech/tags/smolvla.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [state](<https://devfeed.tech/tags/state.md>), [supervisor](<https://devfeed.tech/tags/supervisor.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [time](<https://devfeed.tech/tags/time.md>), [training](<https://devfeed.tech/tags/training.md>), [uncertainty](<https://devfeed.tech/tags/uncertainty.md>), [variance](<https://devfeed.tech/tags/variance.md>), [vision-language-action](<https://devfeed.tech/tags/vision-language-action.md>)

### AI overview

The paper proposes post-hoc kernel-variance heads that estimate epistemic uncertainty from a single VLA policy's embeddings, using a Gaussian process conditioned on training demonstrations. Fitting requires neither regression targets nor gradient updates and leaves the policy unchanged. In SmolVLA experiments on LIBERO-10, the heads performed competitively with two-policy ensemble scores for task-success correlation, active-learning demonstration acquisition, and referral to a stronger supervisor. The vision-language backbone head cut uncertainty-scoring time by up to about 50-fold versus ensemble scoring and reduced setup costs. The authors caution that LIBERO-10's standard protocol mainly rewards recognizing task coverage differences, while execution variability obscures within-task rankings. Controlled perturbations complement that evaluation: modality-specific heads detected unfamiliar camera, instruction, or state inputs, and fused heads detected unfamiliar combinations. With a supervisor proxy trained on those perturbation families, the decoder head improved within-task referral over ensemble disagreement.

### Source excerpt

arXiv:2610.29730v1 Announce Type: cross Abstract: Vision-language-action (VLA) policies need to recognize gaps in their training knowledge to request useful demonstrations or assistance in new situations, especially when failures are costly. Disagreement within an ensemble of independently trained policies can guide these decisions, but requires training and running multiple policies. We propose post-hoc kernel-variance heads that estimate epistemic uncertainty from the embeddings of a single policy. Each head computes the conditional variance of a Gaussian process conditioned on embeddings from the training demonstrations. Fitting requires no regression targets or gradient updates and leaves the policy unchanged. In experiments with SmolVLA on LIBERO-10, the kernel variance heads perform competitively with two-policy ensemble scores in task-level correlation with task success, in demonstration acquisition during active learning, and in referral to a stronger supervisor policy as a proxy for human assistance. For SmolVLA, the head on the vision-language backbone reduces uncertainty-scoring time by up to approximately a factor of 50 compared with ensemble scores, while also substantially reducing setup costs. We also show that the established evaluation protocol on LIBERO-10 primarily rewards recognition of coverage differences between tasks, while execution variability weakens assessment of within-task rankings. Thus, controlled perturbations provide a complementary evaluation: modality-specific heads identify unfamiliar camera, instruction, or state inputs, while fused heads can detect unfamiliar combinations of individually familiar inputs. With a supervisor proxy trained on the evaluated perturbation families, the decoder head improves within-task referral over ensemble disagreement. These findings show that useful uncertainty estimates for adaptive robot learning can be obtained from a single policy.