Robotics & Embodied AI - Weekly Roundup

Embodied agents improve on complex tasks by recombining learned skills, but their ability to extract value from additional experience saturates, failing to scale with learning samples.

Existing 3D asset generators can now produce objects that not only look good but also move and function realistically, passing physics tests for stability and collision-free articulation.

Innate robot reflexes, designed for solitary tasks, actively impede the evolution of effective communication strategies in social foraging.

One million parallel environments are insufficient without intelligent sampling: Success Guided Sampling unlocks mega-scale RL for robots by focusing training on the edge of learned capabilities.

Abstracting human demonstrations into scene graphs unlocks zero-shot generalization for dexterous manipulation policies, achieving 71% better performance on unseen tasks than imitation-heavy methods.

Embodied agents improve on complex tasks by recombining learned skills, but their ability to extract value from additional experience saturates, failing to scale with learning samples.

Existing 3D asset generators can now produce objects that not only look good but also move and function realistically, passing physics tests for stability and collision-free articulation.

Innate robot reflexes, designed for solitary tasks, actively impede the evolution of effective communication strategies in social foraging.

One million parallel environments are insufficient without intelligent sampling: Success Guided Sampling unlocks mega-scale RL for robots by focusing training on the edge of learned capabilities.

Abstracting human demonstrations into scene graphs unlocks zero-shot generalization for dexterous manipulation policies, achieving 71% better performance on unseen tasks than imitation-heavy methods.

Frozen VLAs can recover from unseen actuator faults by inferring joint capabilities online, boosting success rates from 24.8% to 59.3% on 28 tasks.

Pretrained LLMs can drive complex robot navigation tasks across diverse environments and requests without any navigation-specific fine-tuning, by offloading motion execution to specialized tools.

Robot world models can now achieve 6.25% interaction defect rates, a 7-fold reduction, by training on counterfactual futures and a novel embodied video reward.

Robots can achieve 70% success on complex bimanual tasks using only code for decision-making, bypassing runtime vision-language models entirely.

Visual-native skill cards, directly accessible to VLMs, unlock a self-reinforcing loop of policy improvement and skill accumulation, drastically boosting sample efficiency.

A single filter can reduce unsafe motion generation by 90% by grounding safety rules in the generator's own reference trajectories, without retraining.

A single feedforward network learns to predict safety values and amortize minimal-intervention filtering, enabling safe RL in high-dimensional systems without analytical safety functions or competing policy objectives.

A reconstruction-free JEPA successfully disentangles invariant and variant latent factors, offering a path to robust world models without observational decoders.

Manual assembly instruction writing is now obsolete: LLMs can generate accurate, CAD-grounded procedures, preserving engineering details and passing automated checks.

Dynamic critic pessimism, guided by geometric disagreement, unlocks superior locomotion learning and efficiency in off-policy actor-critic methods.

Generated robot futures can actively harm performance if temporally misaligned, but a new control-based conditioning method recovers substantial success by learning to trust and align them.

Human motion alone, when translated into a shared latent space, unlocks zero-shot generalist humanoid control for locomotion and manipulation tasks.

Local contact reasoning, not global scene modeling, unlocks 30%+ accuracy gains in rigid-body physics simulation.

Human tactile gloves costing under $11 can now generate robot-ready tactile data, slashing acquisition costs by over 95% and boosting efficiency 3.5x for dexterous manipulation.

Fine-grained embodied interactions between multiple agents can be effectively modeled by synchronizing ego-stream generation with cross-view action consistency and shared environment memory.

Realistic human-object interactions in complex 3D scenes are now achievable by decoupling scene feasibility from motion synthesis, leveraging separate supervision signals.

Humanoid robots can now catch thrown objects with 65% success from 2.5m using only egocentric RGB-D, a leap in reactive whole-body control.

Latent world models scale predictably with compute via a second-order power law, enabling accurate forecasting of robotic planning performance and zero-shot task execution on real hardware.

A single-word instruction change can tank VLA performance by 98%, but a simple LLM-based rephrasing pipeline boosts success by up to 27% without retraining.

Current multimodal agents can build sophisticated perception-control pipelines but fail to reliably execute physical tasks, losing object states and mistaking unfinished goals for completion.

Autoregressive video pretraining unlocks the full potential of long contexts for real-time robot control, boosting success rates by 15% and enabling complex manipulation previously impossible.

Fingertip touch dynamics, not just vision, unlock a 10.5% boost in dexterous robotic grasp success.

Quadrupedal robots can achieve up to 20% lower energy consumption by learning speed-dependent gaits with a hierarchical RL policy that explicitly optimizes cost of transport.

Physical robot learning can now autonomously discover and exploit informative experiments to continuously improve foundation models, achieving 77% success on RoboCasa365 and zero-shot transfer to real hardware.

Frontier multimodal agents struggle to autonomously acquire and utilize task-relevant information in unfamiliar physical environments, failing 77% of the time on a new exploration benchmark.

Embodied agents can now remember objects they never touched by tracking their resting locations and inferring moves from repeated evidence, boosting downstream task accuracy by over 10%.

Co-evolved communication protocols exhibit surprising robustness when transferred from 2D to 3D physics, with performance scaling linearly with episode length and social cues acting predominantly as jam-assistance.

A single 500-hour visual-tactile dataset boosts zero-shot contact prediction from 13% to 38% IoU and quadruples robot manipulation success rates, proving scale and consistency are paramount for embodied learning.

A simulation flywheel recursively bootstraps embodied data, generating richer scenes and tasks for robot pretraining and evaluation.

Most video world models fail basic physics tests, with even the best models exhibiting significant physical inconsistencies across mechanics, optics, and fluids.

Humanoid locomotion learned from scratch and transferred to hardware in 10x less time, thanks to a novel off-policy flow RL approach.

Embodied agents can now reliably achieve complex, multi-stage goals in dynamic environments by learning to predict and ground targets from decoupled visual context, even when the goal is out of view.

A 48.9M parameter robot policy matches systems 163x larger by prioritizing fine-grained, compressed visual representations over scale or generative priors.

End-to-end driving models can learn interpretable safety costs over reachable futures, drastically reducing collisions without fine-tuning.

A single RL policy masters energy-efficient path following across diverse tracks and even complex vehicle dynamics, outperforming NMPC and baselines without explicit track-specific training.

Robot world models can now predict photorealistic RGB *and* accurate metric depth, bridging the gap between visually plausible rollouts and geometrically consistent 3D environments.

Language instructions now directly control real-time, high-fidelity 360° video generation, enabling fluid interactive exploration from a single panorama.

Accelerating VLA models without sacrificing task success is now certifiable, guaranteeing over $85\%$ of reference performance with $10\times$ speedups.

Explicitly learning a shared perception-action representation boosts continuous-action VLA performance by over 13% on real-world tasks.

Predicting subskill transition timing unlocks reliable execution of 4x longer action chunks in VLA models, drastically cutting robotic idle time by 5x.

Achieve 25x speedup in World Action Models with <1% performance drop by enabling one-step action generation and asynchronous expert execution.

Dual-arm robots can generalize to unseen task compositions by combining skill-conditioned parameters and arm-wise attention, achieving 39% success on physical hardware.

Hierarchical JEPAs unlock efficient long-horizon visual planning by learning distinct latent spaces for different temporal scales, dramatically boosting success rates and reducing compute.

Humanoid robots can now learn complex loco-manipulation tasks from sparse demonstrations by iteratively evolving their own training data and policies.

Asynchronous execution fundamentally alters action semantics for world-model controllers, causing planning models to hallucinate future actions and recurrent models to misattribute transitions.

TSFMs for predictive control fail to capture system dynamics without explicit control excitation in their inference context, even with low forecasting error.

A dual VAE architecture effectively bridges the sim-to-real gap for robotic navigation, enabling high performance with minimal real-world data.

Robots can learn to sacrifice immediate gains for superior team coordination, outperforming centralized and reactive planners even with over a thousand agents.

A single, static adversarial patch can cripple world action models across diverse tasks and architectures by inducing a stable, global representation shift, even without querying the target policy.

Robots can now learn complex manipulation tasks from a single demonstration video by recursively accessing fine-grained sub-event details, boosting success rates by up to 9%.

Simulation-based motion priors can be effectively distilled into real-world robot video generation, overcoming appearance gaps and boosting downstream policy performance.