DexCompose: Role-aware residual framework for composing dexterous manipulation policies with a single hand
Researchers introduce DexCompose, a framework for reusing pretrained dexterous manipulation policies to perform multiple tasks simultaneously with a single robotic hand. The approach assigns explicit finger-level action ownership and trains asymmetric residual modules — a bounded stabilizer for preserving existing skill states and a context-aware residual for new task execution — to avoid destructive interference between overlapping skills. Evaluated on 16 composite manipulation tasks, DexCompose achieves a 77.4% average composite success rate, outperforming conventional policy chaining approaches.
Related guides (1)
Related events (8)
CoorDex: Learning pipeline for continuous dexterous humanoid loco-manipulation with high-DoF hands
CoorDex is a reinforcement learning pipeline that enables humanoid robots to perform dexterous manipulation while walking, eliminating the stop-and-go pattern common in prior work. The approach trains separate privileged motion tracking teachers for body and hand, distills them into latent priors, and uses coordinated residual RL to compose them for downstream tasks. Demonstrated on a Unitree G1 humanoid with a 20-DoF WUJI hand, the system achieves non-stop bottle grasping, fridge door opening, and cube manipulation in motion. Ablations show that naive joint-space or monolithic approaches fail under the same reward budget, validating the latent-prior architecture.
DexHoldem: A Real-World Benchmark for Dexterous Embodied Agents Using Texas Hold'em Manipulation
DexHoldem is a new system-level benchmark for evaluating dexterous embodied agents on a ShadowHand robot performing Texas Hold'em card manipulation tasks. It provides 1,470 teleoperated demonstrations across 14 manipulation primitives, a physical policy benchmark, and an agentic perception benchmark for structured game-state recovery. Top performers include π₀.₅ at 61.2% task completion and Claude Opus 4.7 at 34.3% strict perception accuracy, with GPT 5.5 achieving 66.8% field-wise accuracy. The benchmark exposes gaps between isolated visual sub-capabilities and full closed-loop embodied decision-making.
AutoDex: Automated real-world dexterous grasping data collection with 4.8x throughput over teleoperation
AutoDex is an automated system for collecting physically-labeled dexterous grasping data at scale, closing the perception-execution-labeling-reset loop without human intervention. The system uses 20-camera dense perception to handle hand-object occlusion, executes collision-monitored motions, and actively resets objects between trials. Across 100 objects and two robot hand platforms, AutoDex achieves 4.8x throughput versus teleoperation and yields 76% grasp success from its validated database versus 34% for simulation-only validation. Code and data will be publicly released.
Mana framework achieves zero-shot sim-to-real transfer for dexterous articulated tool manipulation
Researchers introduce Mana (Manipulation Animator), a sim-to-real framework that reframes dexterous robotic manipulation as an animation problem using a coarse-to-fine pipeline of procedurally-generated grasp keyframes, motion planning, and reinforcement learning. The system requires minimal human input (under one minute per tool) and achieves zero-shot sim-to-real transfer across four articulated tools with varying joint types and scales. The work addresses a longstanding gap in dexterous robotics where articulated tool use—requiring coordination of internal degrees of freedom and contact-rich interactions—has been underexplored relative to rigid object manipulation.
Beyond Binary: Sim-to-Real Dexterous Manipulation with Physics-Grounded Contact Representation (CoP)
Researchers introduce Center-of-Pressure (CoP), a tactile representation grounded in physical principles designed to bridge the sim-to-real gap in contact-rich dexterous manipulation. CoP preserves dense contact information while remaining robust for sim-to-real transfer, supported by a differentiable-dynamics-based sensor calibration scheme that estimates taxel orientations without ground-truth force measurements. Evaluated on peg-in-hole insertion and ball balancing tasks, CoP-conditioned policies achieve zero-shot sim-to-real transfer on a multi-fingered robotic hand, outperforming binary-contact and raw-taxel baselines. An emergent finding is that CoP-conditioned policies implicitly encode task-relevant physical properties such as object mass.
SkillComposer: Structured skill composition for LLM agents via constrained autoregressive decoding
A new arXiv preprint introduces SkillComposer, a method that frames skill selection for LLM agents as a structured prediction problem — jointly deciding which skills to activate, how many, and in what order via a constrained autoregressive decoder over skill identifiers. The approach addresses a bottleneck in growing skill libraries where existing retrieval and full-context methods fail to capture the joint nature of skill composition. Evaluated on SkillsBench across two production-grade coding agents (GPT-5.2-Codex and Gemini-3-Pro-Preview), SkillComposer raises pass rates by +23.1 and +18.2 percentage points over no-skill baselines, matching gold-skill retrieval upper bounds at lower prompt-token cost.
Learning Dexterity: OpenAI Trains Robot Hand for Physical Object Manipulation
OpenAI announced the training of a human-like robot hand capable of manipulating physical objects with what they describe as unprecedented dexterity. The system uses reinforcement learning to develop fine motor control in a dexterous robotic hand. This work represents an early milestone in OpenAI's robotics research program, predating their later Dactyl work on solving Rubik's cubes.
SkillWeaver: Compositional Skill Routing for LLM Agents via Decompose-Retrieve-Compose
Researchers introduce SkillWeaver, a framework for compositional skill routing in LLM agents that decomposes complex queries into atomic sub-tasks, retrieves matching skills from a large library, and composes an executable DAG plan. The paper formalizes the Compositional Skill Routing problem and introduces CompSkillBench, a benchmark of 300 compositional queries over 2,209 real MCP server skills across 24 categories. A key finding is that task decomposition quality is the primary bottleneck, with standard LLM decomposition reaching only 34.2% category recall; the proposed Iterative Skill-Aware Decomposition (SAD) method improves decomposition accuracy from 51.0% to 67.7% in a single iteration. The framework also reduces context window consumption by over 99% compared to naive skill-stuffing approaches.