All papers

The paper proposes a vision-language-action model that dynamically crops high-resolution regions to manipulate objects without using a wrist camera.

The paper presents a framework adapting video generation priors to improve 3D human body and hand motion reconstruction from video.

The paper introduces a dataset and framework pairing synchronized point-of-view and panoramic video to evaluate multiplayer action-conditioned world models.

The paper detects backdoor triggers in multimodal language models by tracking output evolution trajectories across internal hidden layers.

The paper presents a benchmark dataset evaluating unified page-level document parsing alongside alternative text descriptions for visual elements.

The paper retargets object motion rather than human hands from video demonstrations to generate post-training data for various robot embodiments.

The paper presents a method to consolidate context gains into parametric model weights using prospective queries and low-rank feed-forward updates.

Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding.

Evaluates patient linkage and attribute inference risks present in brain magnetic resonance imaging representations generated by pretrained foundation models.

Uses explicit three-dimensional vascular guidance in an iterative diffusion pipeline to improve anatomical accuracy in computed tomography super-resolution.

Presents a graph-based framework with decision certificates to determine when multimodal agents should acquire visual evidence under cost constraints.

Introduces a decomposed rubric synthesis method to create training data for open multimodal models that generate visual code artifacts.

Proposes an interactive agent framework that utilizes physics simulation and spatial exploration for complex 3D object assembly and composition.

Develops a post-training framework that trains computer-use agents for video games using synthetic interactive experiences and an expanded action space.

Proposes a hierarchical residual policy learning framework to fine-tune mobile robot manipulation policies with decoupled base and arm adjustments.

Generatively models humanoid robot control policy weights using a latent diffusion framework combined with a linear autoencoder and Diffusion Transformer.

This paper proposes a discrete diffusion model for controllable visual layout generation under structural element constraints.

This paper introduces an evaluation benchmark using multimodal models to assess generated videos across human preference dimensions.

This paper analyzes the accuracy limits of pose-graph optimization compared to bundle adjustment for drone-based Gaussian splatting.

This paper evaluates a multimodal neural network that fuses drone camera images and kinematic data for communications beam prediction.

This paper extends split Gibbs sampling to non-Euclidean geometries using Jensen-Bregman divergence for imaging inverse problems.

This paper introduces a flow-matching method that preserves regional probability mass to model rare events more accurately.

This paper proposes an agentic architecture that dynamically coordinates distributed drone capabilities based on workflow demands.

This paper proposes an optical system design framework based on Fisher information to improve downstream image restoration performance.

The paper adds a lightweight decision head to a robot policy to call an external reasoning model only during critical recovery events.

Introduces GPU-parallel adaptive mesh refinement to simulate deformable object manipulation efficiently by focusing resolution on contact areas.

Presents a method to forecast surgical tissue deformation from preplanned instrument trajectories using observation-referenced dynamic models.

Proposes an offline smoothing method to estimate and map tire-road friction coefficients from standard on-board vehicle sensors.

Presents a dynamics compiler extension that models coupled multi-body kinematics and forces for teams of physically connected humanoids.

Introduces an authorization and monitoring framework to verify and bound autonomous excavator motions generated by language models.

Guides long-horizon world-model planning in robotic manipulation using video-generated visual subgoals and learned policy priors.

Combines vision-language models with spatial overlays and chain-of-thought prompting to plan coordinated paths for multi-robot teams.

Presents a residual learning approach for physics-based demonstration retargeting to achieve zero-shot robotic grasping across arbitrary objects.

Introduces quality-diversity optimization to generate adversarial environments for red-teaming and fine-tuning vision-language-action models.

Demonstrates a shared-control teleoperation system enabling continuous industrial valve rotation using omnidirectional aerial robots.

Proposes a two-stage polar gradient field method for real-time vision-based aerial vehicle trajectory planning in dynamic environments.

Develops a fourth-order control barrier function framework incorporating full nonlinear dynamics for safe quadrotor obstacle avoidance.

Presents a decentralized multi-robot manipulation method that predicts teammate responses to coordinate collective actions under local sensing.

Introduces a multi-stage reinforcement learning framework to teach humanoid robots generalizable object carrying directly from video demonstrations.

Proposes an action-filtering framework enabling robots to infer human visual visibility and adapt behavior during human-robot interactions.

Introduces a seven-dimensional deformation model based on similarity transformation groups to simulate expanding and contracting soft robots.

Introduces a learning framework enabling quadruped robots to estimate carried object geometries from contact and adapt navigation paths.

Develops an embodied agent architecture using finite-state machine skill representations to structure and reuse multimodal interaction policies.

Proposes a sequential testing method to reliably rank robotic policies with minimal physical hardware trials.

Presents an asynchronous cloud-edge dual-policy architecture connecting high-level visual guidance to fast local trajectory planning for navigation.

Provides theoretical conditions and explicit noise bounds guaranteeing the global optimality of pose graph optimization in robotics.

Introduces a benchmark evaluating collision safety and blocking dynamics in vision-and-language navigation agents.

Introduces a training-free framework for multi-altitude drone collaboration in search-and-rescue using belief-guided memory coupling.

Proposes a training method for vision-language-action policies to fix supervision imbalance and improve high-precision robotic manipulation.

Presents a co-simulation framework to evaluate heterogeneous computing systems and closed-loop flight performance for drones before building hardware.

The paper introduces CAG-WAM, applying a structural causal model to isolate action-induced improvements in world action models for manipulation.

The paper details B-Human's behavior architecture and system modifications for scaling humanoid soccer play up to 11-versus-11 matches.

The paper introduces TempoLoco, a body-aware dynamic visual perception framework that adapts visual memory to current quadruped locomotion states.

The paper evaluates whether rapid motor adaptation conditioning on latent dynamics scales effectively to coordinated bimanual robotic manipulation tasks.

The paper presents a learning-based framework using multi-head attention to enable high-speed sprinting and aggressive cornering on full-scale humanoid robots.

The paper proposes a conditional flow matching method to synthesize dynamic 4D hand-object grasping sequences from single initial configurations.

The paper presents DEFT-VLA, a dual-expert vision-language-action framework specializing force feedback specifically for contact stages during robotic manipulation.

The paper proposes a contextual masked action policy that performs history-conditioned local span regeneration to refine continuous robotic action chunks.

The paper presents Verify, Act, Distill, distilling inference-time search rollouts into LoRA parameters for vision-language-action models without reinforcement learning.

The paper proposes LETS, a physics-guided residual learning method for real-time external-torque sensing in force-feedback bilateral teleoperation.