One VLA policy places 9th in the 2026 BEHAVIOR Challenge
A single π0.5-based checkpoint scores 0.392 Q on 100 tasks; GPT-6 Astra averages 0.639 Q on 14 routed tasks.
Stories per day · 30 days
8 RSS
Relatedrobotics21embodied-ai8vlm6benchmark3libero3vision-language-action3flow-matching2personalization2reinforcement-learning2robomme2robot-safety2action-atlas1action-chunk1adversarial1adversarial-robustness1agent-safety1atomicrl1autonomous-driving1
71 stories, newest first · one card per story
Stories tagged #vla
Memory VLAs treat pauses as progress via temporal shortcuts
Brief pauses drop counting success from 47 to 16 of 50 and from 44 to 10; event-indexed memory recovers half or more without retraining.
LingBot-Touch learns shared tactile tokens across sensor types
Authors say the tokens plug into diffusion policies and VLA models, including pi0.5 and LingBot-VLA.
VLA-Tug probe improves failure detection in VLAs
VLA-Tug uses latent probing to detect and classify failures in VLAs, improving early detection by 30% and accuracy on unseen tasks by 44%.
Visual perturbations alter VLA actions without changing text rationale
TNAS-guided attacks induce collisions in 52.6% of driving scenes while keeping rationales word-for-word identical.
H2RB bridges human-to-robot VLA transfer
H2RB co-trains VLAs on unpaired human and robot data for 51.8 object transfer success.
Visual perturbations decouple actions from identical rationales in VLAs
Alpamayo and InstructVLA show TNAS steering shifts actions 4.64x-47.8x more than text perturbations while preserving rationales.
REA runtime coordinates slow planners with fast robot policies
On 3,000 LIBERO-Pro episodes it reports 95.9% success, 25% fewer control steps, and 17% less control time than waiting on the brain.
ThinkWorld grounds robot action reasoning in imagined world dynamics
Chain-of-Dynamics tokens let unlabeled human and robot videos supervise the same action space via predicted state transitions.
GPT-6-Astra evaluated on EBench for direct and policy-harness robot control
Direct control leads on short tasks but lags on long-horizon work; calling a VLA or WAM matches top policies at lower token cost.
OSMA-Bench finds robot policies fail obstacle-conditioned safe motion
VLM and VLA policies can rank obstacles by safety but do not change clearance, speed, or smoothness to match.
RynnVLA-Latent pretrains a 4B robot policy on 3D latent actions
A 4B policy pretrained on 745.52M descriptors from 30.1K hours of video posts strong LIBERO, VLABench, and RoboTwin scores.
HuGe pretrains humanoid VLAs on 2,000 hours of egocentric data
The three-stage pipeline reports state-of-the-art simulation results and higher Unitree G1 pick-and-drop success.
Paper sets safety rules for sharing human corrections across robots
Authors want context preserved, closed-loop checks, and leakage measured, but report no experiments.
Sensorimotor Alignment proposed as offline proxy for robot policy success
The training-free S^2 metric correlates with closed-loop robot task success and can rank VLM backbones before VLA fine-tuning.
New offline metric measures sensorimotor alignment in AI models
Sensorimotor Alignment S2 correlates with robot task success and identifies strong vision-language-action models before expensive rollouts.
AtomicRL lifts long-horizon VLA robot performance over PPO
Reported gains are 3.6–4.0% on LIBERO-LONG, up to 9.5% on CALVIN, and 21.3% on real robots.
SERA self-evolving harness lifts long-horizon robot planning on RoboMME
With a fixed VLA executor, autonomous harness selection reached 51.38% success versus 44.38% without self-evolution.
RoboCasa-Shift measures stage-aware robustness of household VLA policies
GR00T N1.5 ran 1,040 rollouts; equal success drops hid different first-stage failure rates.
Activation study finds VLA motor programs ignore shifted scenes
Injected task-A activations made pi0.5 replay that trajectory in 99.6% of 1,968 episodes.
Meet Kyutai REKI: The 2.4B VLA Model Bringing Efficient Memory to Robot Manipulation
Kyutai introduced REKI, a 2.4B vision-language-action model for memory-efficient robot manipulation.
VLA-Rolling reuses action tails to cut VLA denoising steps
Across 16 frozen model-benchmark pairs it lowered chunk inference cost, with up to 1.48 times speedup and higher success in 14.
NestedVLA: Learning to Consolidate and Generate Skills for Vision-Language-Action Models
NestedVLA generates task-specific VLA parameters, beating model merging by 41 points across three robotics benchmarks.
RoboHarness lets an LLM agent beat VLA models on robot tasks
The open-source harness gives agents a visual-geometric control panel; Qwen3.8-Flash-Next leads BEHAVIOR Challenge 2025.