The Archive

PolicyIK uses RL to learn high-precision inverse kinematics for robot arms, enabling transfer across embodiments in vision-language-action policies.

AceAction-5M: 5M-scale vision-language-action dataset from 3D games with hierarchical goals for generalist embodied learning.

TinyLoRA adapters with 256–384 parameters enable behavioral steering of vision-language-action models while retaining pretrained capability.

BURGER benchmark evaluates vision-language-action model generalization via scalable prompt-driven robotic drawing in simulation and hardware.

Vision-Language-Action models trained on 1616 underwater manipulation episodes for robotic task generalization.

Martingale GRPO refines vision-language-action RL by applying trajectory-level penalties only to actions causally linked to failure, not recovery attempts.

SP-HLM couples sparse-pose interface with learned motion priors and vision-language-action models for humanoid loco-manipulation control.

NestedVLA consolidates multi-task robotic skills via hypernetwork-based parameter generation instead of parameter merging.

AdaRSI framework selectively retrains vision-language-action model executors via on-demand recursive self-improvement for long-horizon embodied tasks.

LIBERO-RIG benchmark evaluates vision-language-action robot policies across 11 hardware perturbation axes including camera calibration and gripper wear.

Vision-language-action models show severe instruction phrasing sensitivity; rephrase augmentation mitigates but doesn't eliminate performance variance across phrasings.

DRIVE uses diversity-driven RL fine-tuning to improve vision-language-action model generalization under distribution shift.

EgoLAP pre-trains vision-language-action models on human+robot trajectories via language-based motion intent chains.

Post-training quantization degrades robustness of vision-language-action models; vulnerability-oriented tuning recovers generalization under distribution shift.

PermVLA: factorization order as regularizer for VLA policies; causally anchored permutation improves vision-language-action learning.

MemBodied adds fixed-size episodic memory to vision-language-action policies for history-dependent manipulation; retains past observations without context bloat.

LADA pipeline reduces language annotation bottleneck for vision-language-action models by learning latent action representations from unlabeled trajectories.

Cross-policy analysis of 15k LIBERO rollouts reveals vision-language-action policies achieve similar end-effector geometry when succeeding.

HIL-UMI: human-in-the-loop post-training for vision-language-action robot models on Universal Manipulation Interface via interactive policy refinement.

rMuscle optimizes Vision-Language-Action model inference latency for factory robotics by exploiting embodied workload characteristics.

ActiveScale framework enables vision-language-action models to perform active perception via coordinated model, data, and hardware designs for robotic manipulation.

ActionPiece rethinks action tokenization in vision-language-action models to preserve task-relevant adjustments beyond MSE.

WetRobo is a reproducible robot kit for wet-lab automation that transfers vision-language-action policies between laboratory environments without teleoperation.

Dynin-Robotics omnimodal diffusion model unifies language-conditioned action, goal, and dynamics prediction for robot control.

TANGO: vision-language-action model for humanoid robot navigation in cluttered indoor spaces with whole-body adaptation.

DeCAL integrates contact-aware latent co-imagination into vision-language-action models to handle dexterous manipulation with severe visual occlusions and contact dynamics.

RoboSPA benchmark evaluates Vision-Language-Action models on spatial reasoning and procedural complexity in robotic manipulation tasks.

DroneCATS benchmark evaluates MLLMs as generalist vision-language-action agents for drone control with full action-space prompting.

LM-X: vision-language-action robot policy with explicit task-progress, event, and uncertainty supervision for long-horizon manipulation.

Intention Distillation method improves Vision-Language-Action models by supervising shared semantic objectives of robot behavior beyond behavior cloning.