mlsanityData to 14 Oct 2026

Drift is a Sampling Error: SNR-Aware Power Distributions for Long-Horizon Robotic Planning

arXiv 2605.09537 ↗First posted 10 May 2026cs.ROcited by 3

On the map: Vision-language-action models → Reinforcement learning and post-training

Abstract (arXiv)

Despite rapid progress in Vision-Language-Action (VLA) models for robotic control, instruction drift remains a persistent failure mode in long-horizon tasks. This paper reconceptualizes this phenomenon, positing that instruction drift is fundamentally a systematic sampling error: local greedy sampling is prone to collapsing into "Negative Pivotal Windows"--irreversible local optima with high local probability that sever global success pathways. To address this, we propose Context-Aware Power Sampling (CAPS), a training-free inference-time computation framework. CAPS leverages power distributions to sharpen global trajectory probabilities, enabling lookahead search over the model's conditional generative trajectory distribution. Furthermore, we introduce a metacognitive control mechanism based on Signal-to-Noise Ratio (SNR). This mechanism triggers adaptive MCMC search solely when drift risk is detected, enabling a dynamic transition from "intuitive fast thinking" to "rational slow search." Experiments on RoboTwin, Simpler-WindowX, and Libero-long benchmarks show that CAPS achieves substantial improvements over strong baselines, including OpenVLA and TACO, without parameter updates. These results support the effectiveness of adaptive inference-time computation for improving long-horizon robustness in embodied control.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Reinforcement learning and post-training, one of 13 directions in Vision-language-action models. Builds on 32 of its 40 references that are in the map. First posted on arXiv 10 May 2026.

Builds on

32 of 40 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Octo: An Open-Source Generalist Robot Policy20 May

Other references · 31

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  3. 2023Let's Verify Step by Stepcited by 3,789
  4. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,211
  5. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 3,080
  6. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  7. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  8. 2022Do As I Can, Not As I Say: Grounding Language in Robotic Affordancescited by 2,348
  9. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,761
  10. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  11. 2023Open X-Embodiment: Robotic Learning Datasets and RT-X Modelscited by 862
  12. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 691
19 more
  1. 2024BLINK: Multimodal Large Language Models Can See but Not Perceivecited by 550
  2. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 519
  3. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 495
  4. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 451
  5. 2024Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Modelscited by 290
  6. 2024VLABench: A Large-Scale Benchmark for Language-Conditioned Robotics Manipulation with Long-Horizon Reasoning Taskscited by 175
  7. 2025Uncertainty Quantification and Confidence Calibration in Large Language Models: A Surveycited by 95
  8. 2025Reasoning with Sampling: Your Base Model is Smarter Than You Thinkcited by 89
  9. 2024The Ingredients for Robotic Diffusion Transformerscited by 73
  10. 2024M2Diffuser: Diffusion-based Trajectory Optimization for Mobile Manipulation in 3D Scenescited by 27
  11. 2025The Pitfalls of Imitation Learning when Actions are Continuouscited by 26
  12. 2024Long-horizon Locomotion and Manipulation on a Quadrupedal Robot with Large Language Modelscited by 21
  13. 2025Steering Vision-Language-Action Models as Anti-Exploration: A Test-Time Scaling Approachcited by 21
  14. 2025A Survey of Slow Thinking-based Reasoning LLMs using Reinforced Learning and Inference-time Scaling Lawcited by 19
  15. 2024QUEST: Quality-Aware Metropolis-Hastings Sampling for Machine Translationcited by 16
  16. 2025REALM: A Real-to-Sim Validated Benchmark for Generalization in Robotic Manipulationcited by 15
  17. 2024Enhancing Multi-Robot Semantic Navigation Through Multimodal Chain-of-Thought Score Collaborationcited by 14
  18. 2025VLDrive: Vision-Augmented Lightweight MLLMs for Efficient Language-grounded Autonomous Drivingcited by 6
  19. 2025PRISM: Pointcloud Reintegrated Inference via Segmentation and Cross-attention for Manipulationcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026CANVAS: Consistency-Aware Navigation via Visual Adaptive Sampling for Long-Context Text-to-SVG Generationcited by 0
  2. 2026ReSync: Re-Aligning the Two Clocks of Asynchronous World-Action Modelscited by 0
  3. 2026ReSight-SMC: Two-Stage Power Sampling via Island SMC with Visual Scoutscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026VLS: Steering Pretrained Robot Policies via Vision-Language Models
  2. 2025Exploration and Adaptation in Non-Stationary Tasks with Diffusion Policies
  3. 2025SITCOM: Scaling Inference-Time COMpute for VLAs
  4. 2025Verifier-free Test-Time Sampling for Vision-Language-Action Models
  5. 2026Retrieve-then-Steer: Online Success Memory for Test-Time Adaptation of Generative VLAs
  6. 2025TRACE: A Self-Improving Framework for Robot Behavior Forecasting with Vision-Language Models
  7. 2026Seeing Farther and Smarter: Value-Guided Multi-Path Reflection for VLM Policy Optimization
  8. 2026ELASTIC: Efficiently Learning to Adaptively Scale Test-Time Compute for Generative Control Policies