3505 papers
The paper introduces REVEL, a new task for streaming, fine-grained interactive video manipulation on any object at any time, and proposes DragStream, a training-free method that resolves latent distribution drift and context interference in autoregressive video diffusion models through adaptive distribution self-rectification and spatial-frequency selective optimization.
This paper presents a state-of-the-art speaker verification system leveraging the large-scale w2v-BERT 2.0 model with MFA structure and LoRA fine-tuning, which achieves top performance on Vox1-O and Vox1-H benchmarks while utilizing knowledge distillation-guided structured pruning to reduce model size by 80% with minimal accuracy loss.
This paper introduces PAD-TRO, a novel direct trajectory optimization framework that integrates a gradient-free projection mechanism into the reverse diffusion process to generate dynamically feasible state sequences, achieving zero dynamic feasibility errors and a significantly higher success rate in complex quadrotor navigation compared to existing single-shooting approaches.
This paper introduces tokenizers as a novel and effective attack vector for membership inference against large language models, demonstrating their significant privacy leakage risks through extensive experiments and proposing an adaptive defense to mitigate these vulnerabilities.
This paper presents and validates a novel robotic framework for automated pollination in controlled environments that integrates 3D vision-guided stem grasping with physics-based vibration modeling to safely and effectively induce pollen release with a 92.5% success rate.
This paper introduces "Differentiable Variable Fonts," a framework that mathematically connects variable font parameters to vector graphics to enable gradient-based optimization for automated, intuitive text design and animation workflows while preserving legibility and aesthetics.
This paper introduces Emerging-Barrier Model-Based Diffusion (EB-MBD), a novel approach that integrates progressively introduced barrier functions inspired by interior point methods to overcome the sample inefficiency and catastrophic performance degradation of standard Model-Based Diffusion in highly constrained environments, achieving superior solution quality and computational efficiency without expensive projection operations.
The paper introduces AR-Drag, a reinforcement learning-enhanced few-step autoregressive video diffusion model that achieves real-time, high-fidelity image-to-video generation with diverse motion control while significantly reducing latency compared to existing bidirectional approaches.
This paper proposes Concept-Driven Exploration (CDE), a reinforcement learning framework that leverages a pre-trained vision-language model to generate object-centric concepts as noisy supervisory signals, using concept reconstruction accuracy as an intrinsic reward to guide efficient, targeted exploration in visual control tasks and achieve robust real-world transfer.
This paper investigates how different deliberation protocols (synchronous vs. round-robin) and model architectures influence value alignment and verdict revision in multi-turn LLM debates, revealing significant behavioral disparities where GPT-4.1 exhibits strong inertia and autonomy-focused reasoning while Claude 3.7 Sonnet and Gemini 2.0 Flash demonstrate greater flexibility, empathy, and susceptibility to order effects.
This paper proposes a lightweight, training-free plugin called Functional Head Identification and Class-Conditioned Rescaling that mitigates multimodal hallucinations in large reasoning models by adaptively rebalancing perception and reasoning contributions across layers, achieving significant performance gains with minimal computational overhead.
This paper proposes a preference-conditioned multi-objective reinforcement learning framework that enables a single humanoid locomotion policy to dynamically balance accurate command tracking with compliant responses to external forces, validated through stable training and successful deployment in both simulation and real-world experiments.
This paper introduces DropVLA, an action-level backdoor attack that covertly manipulates Vision-Language-Action models to execute specific safety-critical actions at attacker-chosen decision points using minimal vision-based data poisoning while maintaining high nominal task performance.
This paper presents a demonstration-free framework that combines a learned ego-vision world model with sampling-based Model Predictive Control and a surrogate value function to enable humanoid robots to perform robust, real-time physical contact planning in unstructured environments.
This study found that a protégé-based approach, where individuals teach others about managing digital stress, did not significantly reduce their own problematic digital behaviors compared to control groups, highlighting the challenges of translating cognitive engagement into actual behavioral change.
This systematic scoping review synthesizes thirty-three studies on unsupervised deep generative models for neuroimaging anomaly detection, highlighting their potential for pathology-agnostic localization in data-scarce settings while identifying key challenges such as methodological heterogeneity and limited external validation.
This paper introduces a robust, model-free probabilistic metric that evaluates 6D placement poses from partial point clouds by jointly scoring stability, graspability, and clearance, thereby enabling reliable and unified pick-and-place reasoning for unseen objects on diverse support geometries.
This paper introduces the Continual Audio-Visual Segmentation (CAVS) task and proposes a Collision-based Multi-modal Rehearsal (CMR) framework that effectively addresses multi-modal semantic drift and co-occurrence confusion through novel sample selection and frequency adjustment strategies, significantly outperforming existing single-modal continual learning methods.
This paper introduces PolyJailbreak, a novel black-box framework that exploits multimodal safety asymmetries through a structured library of atomic strategies and reinforcement learning-based multi-agent optimization to achieve significantly higher jailbreak success rates on state-of-the-art multimodal large language models compared to existing methods.
This paper presents HumanMPC, a Model Predictive Control framework that ensures safe and efficient 3D navigation for Micro Air Vehicles among humans by combining data-driven motion forecasting with a novel reachability-based safety formulation that minimizes conservatism while guaranteeing collision avoidance.