Papers I read

Gemma 2: Improving Open Language Models at a Practical Size

Language model

The Gemma 2 open language model report.

Sigmoid Loss for Language Image Pre-Training

Vision-language

SigLIP.

PaliGemma: A versatile 3B VLM for transfer

Vision-language

A transferable 3B vision-language model.

OpenVLA: An Open-Source Vision-Language-Action Model

Vision-language-action

An open VLA model for robot control.

Octo: An Open-Source Generalist Robot Policy

Vision-language-action

An open generalist robot policy.

FAST: Efficient Action Tokenization for Vision-Language-Action Models

Vision-language-action

Action tokenization for VLA models.

π0: A Vision-Language-Action Flow Model for General Robot Control

Vision-language-action

A flow-matching VLA for general robot control.

π0.5: a Vision-Language-Action Model with Open-World Generalization

Vision-language-action

Generalization to new homes through co-training.

πRL: Online RL Fine-tuning for Flow-based Vision-Language-Action Models

Vision-language-action

Online RL for flow-based VLAs.

GR00T N1: An Open Foundation Model for Generalist Humanoid Robots

Vision-language-action

An open foundation model for humanoid robots.

MolmoAct2: Action Reasoning Models for Real-World Deployment

Vision-language-action

Action reasoning models for real robots.

Diffusion Policy: Visuomotor Policy Learning via Action Diffusion

Robot learning

Visuomotor policies learned through action diffusion.

Flow Matching for Generative Modeling

Robot learning

The generative method behind the flow-based policies.

Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware

Robot learning

Bimanual manipulation on low-cost hardware.

A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning

Robot learning

The DAgger paper.

BEHAVIOR-1K: A Benchmark for Embodied AI

Simulation

A benchmark for household tasks in simulation.

OmniGibson: a Simulation Environment for Interactive Tasks

Simulation

The OmniGibson simulator.

Openpi Comet: Competition Solution For 2025 BEHAVIOR Challenge

Simulation

A 2025 BEHAVIOR Challenge solution.

Task adaptation of Vision-Language-Action model: 1st Place Solution for the 2025 BEHAVIOR Challenge

Simulation

The 1st-place solution for the 2025 BEHAVIOR Challenge.

YOLO26

Detection

The Ultralytics real-time detector.

RF-DETR

Detection

A real-time detection transformer.

Open X-Embodiment: Robotic Learning Datasets and RT-X Models

Robot datasets

A large cross-robot dataset and the RT-X models trained on it.

DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset

Robot datasets

A large manipulation dataset collected in the wild.

Ctrl-World: A Controllable Generative World Model for Robot Manipulation

World models

A controllable generative world model for robot manipulation.