Behavior Graphs: Unsupervised Graph Structures for Interpreting Robot Deployment Data
cs.RO, cs.AI, cs.LG
Submitted: 2026-10-30
Updated: 2026-11-05
Comments: 8 pages, 6 figures, 3 tables. Project website: https://behavior-graphs.com/
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- SCIZOR: A Self-Supervised Approach to Data Curation for Large-Scale Imitation Learning
- SARM: Stage-Aware Reward Modeling for Long Horizon Robot Manipulation
- $\pi^{*}_{0.6}$: a VLA That Learns From Experience
- Embodied Red Teaming for Auditing Robotic Foundation Models
- Sparse Autoencoders Reveal Interpretable and Steerable Features in VLA Models
- Event-Grounded Sparse Autoencoders for Vision-Language-Action Policies
- SPARC: Reliable Spatial Annotations from Robot Demonstrations at Scale
- Grounding Language Plans in Demonstrations Through Counterfactual Perturbations
- RoboSubtaskNet: Temporal Sub-task Segmentation for Human-to-Robot Skill Transfer in Real-World Environments
- STRAP: Robot Sub-Trajectory Retrieval for Augmented Policy Learning
- Guiding Data Collection via Factored Scaling Curves
- TOPReward: Token Probabilities as Hidden Zero-Shot Rewards for Robotics
- DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation