Diversity You Can Actually Measure: A Fast, Model-Free Diversity Metric for Robotics Datasets

arXiv:2603.11634 · cs.RO · Submitted 2026-03-12 · Read on arXiv

cs.RO

Submitted: 2026-03-12

Updated: 2026-10-13

Code: https://github.com/huggingface/lerobot

Terminology

• Benchmark — 1× in this paper · explained in 2 episodes
• Generalization — 1× in this paper · explained in 13 episodes

Sources

• CUPID: Curating Data your Robot Loves with Influence Functions
• ChemGAN challenge for drug discovery: can AI reproduce natural chemical diversity?
• $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control
• Curating Demonstrations using Online Experience
• A Primer on the Signature Method in Machine Learning
• Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots
• Eval all, trust a few, do wrong to none: Comparing sentence generation models
• DataMIL: Selecting Data for Robot Imitation Learning with Datamodels
• A Unified and Predictive Measure of Functional Diversity
• The Vendi Score: A Diversity Evaluation Metric for Machine Learning
• Efficient Data Collection for Robotic Manipulation via Compositional Generalization
• Robot Data Curation with Mutual Information Estimators
• $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization
• Towards Diverse Behaviors: A Benchmark for Imitation Learning with Human Demonstrations
• Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning
• The Signature Kernel
• What Matters in Learning from Offline Human Demonstrations for Robot Manipulation
• Vendi Information Gain: An Alternative To Mutual Information For Science And Machine Learning
• Towards Balanced Behavior Cloning from Imbalanced Datasets
• Vendi-RAG: Adaptively Trading-Off Diversity And Quality Significantly Improves Retrieval Augmented Generation With LLMs

Related papers

• FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
• MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
• RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
• ARC: A Reasoning Recipe for Robot Foundation Models
• HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
• APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies