mlsanityData to 7 Oct 2026

SAIL: Test-Time Scaling for In-Context Imitation Learning with VLM

arXiv 2603.08269 ↗First posted 9 Mar 2026cs.RO cs.AIcited by 2

On the map: LLM reasoning → Spatial and embodied reasoning

also in 1 more
Abstract (arXiv)

In-context imitation learning allows robots to acquire skills from demonstrations, yet one-shot trajectory generation remains fragile under environmental variation. We propose SAIL, a framework that reframes robot imitation as an iterative refinement problem capable of scaling with test-time compute. SAIL utilizes Monte Carlo Tree Search, where each node is a complete trajectory and edges correspond to trajectory refinements. The process is guided by three core components: an automated archive of successful trajectories for contextually relevant retrieval, a vision language model-based scoring mechanism for trajectory evaluation, and a step-level feedback that provides trajectory-aligned scores for iterative refinement. Experiments across six diverse manipulation tasks in simulation and real-world validation clearly demonstrate that increasing test-time compute consistently improves success rates, achieving up to 95% on complex tasks. Our results suggest that trajectory-level test-time scaling is a robust path toward more generalizable robotic agents.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning. Builds on 34 of its 38 references that are in the map. First posted on arXiv 9 Mar 2026.

Builds on

34 of 38 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Robo-Dopamine: General Process Reward Modeling for High-Precision Robotic Manipulation29 Dec
  2. 2025ROVER: Recursive Reasoning Over Videos with Vision-Language Models for Embodied Tasks3 Aug
  3. 2024AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation1 Oct
  4. 2024LLM3:Large Language Model-based Task and Motion Planning with Motion Failure Reasoning18 Mar

Other references · 30

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  2. 2023Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detectioncited by 3,274
  3. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  4. 2024SAM 2: Segment Anything in Images and Videoscited by 2,784
  5. 2022Do As I Can, Not As I Say: Grounding Language in Robotic Affordancescited by 2,302
  6. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  7. 2023SayPlan: Grounding Large Language Models using 3D Scene Graphs for Scalable Robot Task Planningcited by 289
  8. 2023Vision-Language Models as Success Detectorscited by 138
  9. 2023RT-Trajectory: Robotic Task Generalization via Hindsight Trajectory Sketchescited by 131
  10. 2023RoboCLIP: One Demonstration is Enough to Learn Robot Policiescited by 128
  11. 2024Vision Language Models are In-Context Value Learnerscited by 84
  12. 2024Keypoint Action Tokens Enable In-Context Imitation Learning in Roboticscited by 71
18 more
  1. 2026LeRobot: An Open-Source Library for End-to-End Robot Learningcited by 70
  2. 2024In-Context Imitation Learning via Next-Token Predictioncited by 65
  3. 2023Language Models as Zero-Shot Trajectory Generatorscited by 59
  4. 2024RAM: Retrieval-Based Affordance Transfer for Generalizable Zero-Shot Robotic Manipulationcited by 55
  5. 2026RoboReward: General-Purpose Vision-Language Reward Models for Roboticscited by 46
  6. 2025SARM: Stage-Aware Reward Modeling for Long Horizon Robot Manipulationcited by 36
  7. 2024In-Context Learning Enables Robot Action Prediction in LLMscited by 25
  8. 2023Kinematic-aware Prompting for Generalizable Articulated Object Manipulation with LLMscited by 19
  9. 2024Trust the PRoC3S: Solving Long-Horizon Robotics Problems with LLMs and Constraint Satisfactioncited by 19
  10. 2025MimicDroid: In-Context Learning for Humanoid Robot Manipulation from Human Play Videoscited by 17
  11. 2025GSWorld: Closed-Loop Photo-Realistic Simulation Suite for Robotic Manipulationcited by 16
  12. 2024REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context in New Environmentscited by 12
  13. 2025Code-as-Symbolic-Planner: Foundation Model-Based Robot Planning via Symbolic Code Generationcited by 10
  14. 2025Guardian: Detecting Robotic Planning and Execution Errors with Vision-Language Modelscited by 8
  15. 2025Prime the search: Using large language models for guiding geometric task and motion planning by warm-starting tree searchcited by 4
  16. 2025TRACE: A Self-Improving Framework for Robot Behavior Forecasting with Vision-Language Modelscited by 2
  17. 2025Trajectory Adaptation using Large Language Modelscited by 1
  18. 2025GELATO: Multi-Instruction Trajectory Reshaping via Geometry-Aware Multiagent-based Orchestrationcited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026What Language Does and What the Evidence Supports: A Functional Role Taxonomy and Evidence Audit of Language Grounding in Embodied Agentscited by 0
  2. 2026Causal-History Test-Time Scaling for Failure Recovery in Autoregressive World-Action Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025SAIL: Faster-than-Demonstration Execution of Imitation Learning Policies
  2. 2025Learning a Thousand Tasks in a Day
  3. 2026RoboTTT: Context Scaling for Robot Policies
  4. 2026E-TTS: A New Embodied Test-Time Scaling Framework for Robotic Manipulation
  5. 2025Robust Offline Imitation Learning Through State-level Trajectory Stitching
  6. 2024Data Scaling Laws in Imitation Learning for Robotic Manipulation
  7. 2025OSVI-WM: One-Shot Visual Imitation for Unseen Tasks using World-Model-Guided Trajectory Generation
  8. 2026Improving Robotic Imitation Learning via Trajectory Standardization