Libero
PulseAugur coverage of Libero — every cluster mentioning Libero across labs, papers, and developer communities, ranked by signal.
- developed World-Action Models 90%
- instance of World-Action Models 90%
- used by World-Action Models 70%
- instance of OpenVLA-OFT 70%
- used by RoboTwin 2.0 70%
- instance of Vision-Language-Action (VLA) models 70%
- instance of LIBERO-Plus 70%
- used by RoboTwin2.0 70%
- used by Vision-Language-Action (VLA) models 70%
- instance of RoboTwin 2.0 50%
- competes with RoboCasa 50%
16 day(s) with sentiment data
-
Reflective VLA improves embodied AI generalization with action consequences
Researchers have introduced Reflective VLA, a novel approach to vision-language-action (VLA) models designed to improve generalization in embodied control tasks. Unlike reactive models that solely rely on current observ…
-
New G3VLA module enhances robot manipulation VLA models with geometric awareness
Researchers have introduced G$^3$VLA, a novel module designed to enhance Vision-Language-Action (VLA) models for robot manipulation. This module addresses the mismatch between 2D image coordinates and the calibrated geo…
-
New FAFM method generates continuous, stable robotic actions
Researchers have developed Frequency-Aware Flow Matching (FAFM), a novel technique to improve robotic action generation by producing continuous and temporally consistent movements. FAFM addresses limitations in existing…
-
New benchmarks and methods improve AI agent uncertainty quantification
Researchers have developed new methods for quantifying uncertainty in AI agents that interact with graphical user interfaces (GUIs) and in vision-language-action models (VLAs) used in robotics. The first study, "Argus,"…
-
New VLA Models Refine Action Plans in Latent Space for Robotics
Researchers have developed new frameworks for Vision-Language-Action (VLA) models to improve robotic manipulation tasks. One approach, PearlVLA, refines action plans within the latent space of a vision-language model to…
-
Robots learn new tasks while retaining old ones with Skill-Compositional Experts
Researchers have developed a new framework called SCE (Skill-Compositional Experts) to address catastrophic forgetting in embodied continual learning for robots. This framework decomposes task demonstrations into reusab…
-
Research finds phase-localized curation ineffective for AI demonstration filtering
A new research paper published on arXiv investigates the effectiveness of phase-localized curation for filtering manipulation demonstrations in reinforcement learning. The study found that applying curation metrics with…
-
New robot policy models enhance action generation and efficiency
Researchers have developed new methods for robot policy learning that improve efficiency and accuracy in action generation. LeaP, a learnable source prior, optimizes the starting point for action generation by condition…
-
Robotics method uses RL to handle missing sensor data
Researchers have developed RL4IL, a novel reinforcement learning approach designed to enhance multimodal imitation learning in robotics, particularly when sensor data is missing. This method utilizes reinforcement learn…
-
New diagnostic shows vision encoder choice depends on VLA backbone scale
A new diagnostic method called frozen-backbone grafting has been developed to evaluate vision encoders for vision-language-action (VLA) policies. This method tests whether an encoder that performs well on a smaller VLA …
-
New EA-WM Framework Enhances Robotic Manipulation with Event Awareness
Researchers have developed EA-WM, a novel event-aware world model designed to improve long-horizon robotic manipulation. This framework enhances existing visual-feature world models by incorporating task-specification-g…
-
MaskWAM model unifies masks for enhanced robotic control
Researchers have developed MaskWAM, a novel object-centric world-action model designed to improve robotic control through video prediction. By integrating masks as both inputs and predictions using a Mixture of Transfor…
-
AI data curation metrics may not improve policy performance
Researchers have found that metrics used to curate training data for AI policies do not necessarily improve the performance of those policies. In experiments on a pick-and-place benchmark, a metric that was highly effec…
-
New TREAD framework uses VLMs to boost robot learning data
Researchers have developed a new framework called TREAD to improve robot learning by augmenting existing datasets. This method uses large Vision-Language Models (VLMs) to generate more diverse and linguistically rich in…
-
GEAR-VLA framework enhances robotic manipulation generalization
Researchers have developed GEAR-VLA, a new framework designed to improve the generalizability of Vision-Language-Action (VLA) models in robotic manipulation tasks. This approach addresses limitations in current VLA mode…
-
Light-WAM model enhances robot manipulation with efficient future prediction
Researchers have developed Light-WAM, a new lightweight model designed for efficient robot manipulation. This model incorporates future video prediction into its training objectives, enabling it to encode temporal struc…
-
ActionMap improves robot policy learning with voxel heatmap
Researchers have developed ActionMap, a novel voxel heatmap action head designed to improve robot policy learning in vision-language-action (VLA) models. This new head replaces the traditional action decoder, predicting…
-
New AI models enhance robot visuomotor control and memory
Researchers have developed new models for robot visuomotor control, focusing on efficient and predictive coordination. CT-VAM, a cerebello-thalamic-inspired model, uses a compact architecture for fast, task-conditioned …
-
OneRobotics launches OneModel 1.7, bridging embodied AI understanding and action
OneRobotics has released its OneModel 1.7 FrontoStria-RL, a new world-action model designed to bridge the gap between understanding environmental changes and executing correct actions in embodied AI. The model utilizes …
-
Flash-WAM achieves 23x faster inference for world-action models
Researchers have developed Flash-WAM, a new framework for world-action models that significantly speeds up inference time. Traditional models require many denoising steps, making real-time control difficult. Flash-WAM e…