libero

🎁 A collection of utilities for LeRobot.

A Python package to use FPGA development tools programmatically.

Instruction blindness in vision-language-action policies: diagnosis and a low-rank data cure. Paper, model, deconfounded datasets, generator, and the measurement battery.

[CoRL 25] Code for FLOWER VLA for finetuning FLOWER on CALVIN and all LIBERO environments

[NeurIPS 2025] Code for BEAST Experiments on CALVIN and LIBERO.

Anchor-Align (arXiv:2607.13429): VLA finetuning that prevents behavior cloning from erasing pretrained VLM representations (catastrophic forgetting) and aligns language with actions. OOD generalization on a physical xArm7, LIBERO-PRO, LIBERO-Plus and CALVIN.

SUREFlow is a lightweight, Mamba-based vision-language-action model with 179M parameters for robot manipulation. Paper: https://arxiv.org/pdf/2607.10504

Backdoor training and evaluation scripts for vision-language-action models on LIBERO.

A simple, efficient, and extensible evaluation framework for robot policies across diverse benchmarks, supporting both local and remote execution

Automated agentic research loop for vision-language-action (VLA) robot policy self-improvement

Independent VLA research notes: OpenVLA / π0 / Spirit paper reviews, LIBERO reproduction, XLeRobot integration. Transitioning from AD motion planning to embodied AI.

Example usage of a set of simulation benchmarks

Robot-agent harness with a CLI and local Web console for LIBERO short evaluation

Provael — red-team open Vision-Language-Action (VLA) robot policies in simulation and report an Attack Success Rate (ASR). Prove it. Prevail.

VSCode dev container for Microchip PolarFire SoC bare-metal development in VS Code. SoftConsole toolchain, Libero, OpenOCD, clangd.

Online reinforcement learning fine tuning for a flow matching vision language action policy, on a single 12GB consumer GPU.

Proofs of concept about FPGA EDA tools.

coming soon

Voice Controlled SO101 (lerobot) robotic arm

在 RTX 3060 6GB 上复现 SmolVLA × LIBERO，并开展低显存 LoRA 微调、多随机种子评测与遗忘控制实验。Low-VRAM SmolVLA × LIBERO reproduction with LoRA adaptation, fixed multi-seed evaluation, and forgetting controls.