Berkeley AI Research (BAIR)

Explicitly grounding the goal object in view unlocks near-expert performance in object-goal navigation, revealing perception rather than policy as the key limitation.

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

Robot skills can be dramatically improved post-training by autonomously generating practice tasks, diagnosing failures with privileged simulator states, and iteratively refining symbolic skills and system prompts, achieving near-perfect success on novel manipulation tasks.

Robots no longer need human intervention to learn from their mistakes: foundation-model-guided exploration over modular skills can autonomously synthesize corrective rollouts that distill directly into low-level policy weights.

Experiments show that SyncWorld can accurately simulate action outcomes in previously unseen settings, and that its capability of simulating rollouts enables test-time policy improvement without training.

Safe-Stop enables humanoid robots to make safer emergency stop decisions by evaluating stoppability in real-time, rather than relying on fixed maneuvers.

Robots can now learn and adapt in real-time, overcoming inference delays that previously stymied reinforcement learning effectiveness.

Action chunking's real power lies in its ability to create implicit ensembles, boosting robustness and generalization beyond traditional policy approaches.

Anchoring visual SLAM to real-world metrics could revolutionize depth estimation, eliminating scale ambiguity and enhancing navigation reliability.

Substantial challenges in task generalization and visuomotor robustness are revealed, highlighting critical gaps in current dexterous manipulation benchmarks.

State-of-the-art Vision-Language Models fall short in real-world robotic applications, revealing critical gaps in their reasoning capabilities.

RoboDojo reveals that integrating simulation and real-world tasks can significantly enhance the evaluation of robot manipulation policies, bridging the gap between theoretical performance and practical deployment.

STL-shaped rewards lead to tighter velocity tracking and more stable training for quadruped locomotion, outperforming traditional hand-crafted methods.

Language prompts can unlock new capabilities in generalist robot policies, enabling them to tackle complex tasks without extensive retraining.

ASPIRE achieves a staggering 31% success rate on unseen long-horizon tasks, compared to just 4% for prior methods, highlighting its superior adaptability and efficiency.

Synthesizing 48,000 interaction trajectories without human input enables a humanoid robot to learn complex loco-manipulation tasks effectively.

Grounded verification in TEXEDO enables humanoid robots to execute complex motions that are both semantically aligned with text prompts and physically feasible.

Coding agents can now autonomously refine robotic manipulation policies to achieve a staggering 99% success rate on complex tasks, revolutionizing real-world robotics.

Human videos can now be transformed into actionable manipulation data for robots, overcoming traditional barriers in hand-object interaction estimation.

SC3-Eval achieves a remarkable 0.929 Pearson correlation in evaluating robot policies, revealing critical insights into their real-world performance.

Playful learning strategies enable robots to acquire skills that boost performance on new tasks by over 20%, transforming how we approach robotic skill acquisition.

VisualClaw slashes API costs by 98% while boosting accuracy, transforming how VLMs can operate in real-time environments.

RHO achieves a 45.0% success rate in robotic tasks, 2.5x higher than the best multi-turn agent, showcasing a breakthrough in real-time control efficiency.

Extracting action signals from 32,041 hours of human video enables CAIP to outperform leading vision encoders in robotic manipulation tasks by over 30%.

Tactile-reactive policies can boost robotic manipulation success rates by over 30% through innovative data collection and a new Mixture-of-Transformers architecture.

Flow Reversal Steering transforms vague human commands into precise robotic actions, achieving up to 95% higher success rates in real-world tasks with minimal training.

Zero-shot sim-to-real transfer for articulated tool manipulation is now achievable with just a few clicks, revolutionizing how robots interact with complex objects.

FTP-1 not only excels on familiar tactile sensors but also achieves unprecedented success on unseen setups, redefining the potential for cross-sensor generalization in robotic manipulation.

State-of-the-art surgical robotics policies can be disrupted by adversarial attacks, leading to a staggering 61% drop in task success rates.

QGF achieves superior performance in reinforcement learning by optimizing policies solely at test time, sidestepping the instability of traditional training methods.

High-quality dense rewards can elevate robotic manipulation success rates from 50% to near perfection, transforming how robots learn from their environments.

StreamForce achieves real-time video generation with physical control, outperforming traditional models in both responsiveness and realism.

LadderMan enables humanoid robots to climb ladders and manipulate objects with unprecedented robustness and adaptability in real-world scenarios.

Reward models trained only on success are fundamentally misaligned with human values, leading to dangerous over-rewarding of poor robot behaviors.

Unlock bimanual robot skills without bimanual robots: MonoDuo lets you train bimanual policies using only a single-arm robot and a human partner.

Curiosity-driven agents can escape local loops in 3D environments by remembering where they've been and building a persistent map of the world.

Camera pose, largely ignored in video LLMs, unlocks significant gains in spatial reasoning and even improves general video QA when used as a lightweight supervisory signal.

Generate diverse, physically plausible, and language-annotated whole-body motion data for humanoid robots at scale with this new interactive web-based pipeline.

Steering worst-case trajectories with an adversarial network and Boltzmann reweighting dynamics ensembles yields a surprisingly stable and efficient approach to robust RL under dynamics uncertainty.

Unlock zero-shot generalization in robot manipulation by generating diverse, affordance-aware training data with 3D generative models and Vision Foundation Models.

Get 3x the imitation learning performance from your robot with just a few extra cameras.

Running robotic manipulation workloads entirely onboard kills robot batteries, but offloading to the cloud tanks accuracy due to network latency, revealing a critical compute placement trade-off.

Forget simulated manipulation—ManipulationNet offers a global infrastructure for benchmarking robots in the real world, complete with standardized hardware and software, to finally measure progress toward general manipulation.

Unlock autonomous driving with YouTube: a new label-free pretraining method learns driving representations directly from unposed in-the-wild videos, outperforming LiDAR baselines with only a single monocular camera.

Humanoid robots can now perform vision-based parkour, chaining together dynamic skills like climbing, vaulting, and rolling, adapting to real-time obstacle changes.