4304 papers
This paper introduces REAP, a novel expert pruning method that outperforms existing merging techniques for compressing Mixture-of-Experts models in generative tasks by leveraging router gate-values and activation norms to minimize reconstruction error, achieving near-lossless compression even at 50% expert reduction.
RL-100 is a unified real-world reinforcement learning framework that combines diffusion visuomotor policies with a clipped PPO objective and consistency distillation to achieve 100% success across 1,000 diverse robotic manipulation trials, matching or surpassing human experts while demonstrating robust zero-shot generalization and continuous deployment in dynamic environments.
FALCON addresses the spatial reasoning limitations of existing 2D-based vision-language-action models by leveraging spatial foundation models to inject rich 3D geometric priors directly into the action head, achieving state-of-the-art performance across diverse simulation and real-world tasks without requiring architectural changes or specialized sensors.
This paper introduces SynHLMA, a novel framework that synthesizes hand manipulation sequences for articulated objects by aligning natural language instructions with a discrete human-object interaction representation, thereby enabling robust grasp generation, prediction, and interpolation for applications in embodied AI and robotics.
The paper proposes GraphKeeper, a novel framework for Graph Domain-Incremental Learning that addresses catastrophic forgetting through knowledge disentanglement and deviation-free preservation, achieving state-of-the-art performance across multiple graph domains while remaining compatible with various graph foundation models.
This paper proposes a structured matrix scaling approach for multi-class calibration that leverages theoretical insights from logistic regression, combined with structured regularization and robust optimization, to effectively manage the bias-variance tradeoff and achieve substantial performance gains over existing methods while providing an open-source implementation.
This paper introduces "information capacity," a novel metric that evaluates large language model efficiency by measuring text compression performance relative to computational complexity while accounting for tokenizer efficiency, thereby enabling accurate performance prediction and revealing linguistic biases across diverse models.
This paper proposes LTSV, a lightweight and efficient method for valuing time series data in foundation models by leveraging in-context finetuning and temporal block aggregation to overcome the computational bottlenecks and temporal dependency limitations of traditional approaches.
This paper introduces MediRound, a new task and baseline model for multi-round entity-level reasoning segmentation in medical images, supported by the large-scale MR-MedSeg dataset and a Judgment & Correction Mechanism to mitigate error propagation in multi-turn medical dialogues.
This paper introduces a novel in-context learning approach using Time-Series Foundation Models (TSFMs) to classify bearing health status from vibration data without fine-tuning, enabling scalable, zero-shot maintenance solutions across varying operational conditions.
This paper presents a Google Gemini 2.0 Flash-based chatbot that automates the setup, execution, and post-processing of two-dimensional finite element eddy current simulations using Gmsh and GetDP, thereby significantly reducing the time required to generate electromagnetic models with customizable geometries and analysis routines.
This paper proposes the Adaptive Diversity Cache (ADC), a novel training-free and plug-and-play module that mitigates long-tail bias in Human-Object Interaction detection by dynamically accumulating diverse high-confidence features during inference to enhance rare category performance without additional training.
This paper proposes "Periodic Asynchrony," a framework that accelerates LLM reinforcement learning by decoupling inference and training into a provably on-policy asynchronous pipeline, achieving a 3- to 5-fold throughput improvement on NPU platforms without sacrificing accuracy.
This paper introduces UPA-RFAS, a unified framework that generates universal and transferable physical adversarial patches to effectively attack diverse Vision-Language-Action (VLA) models across unknown architectures, finetuned variants, and sim-to-real shifts by leveraging robust feature alignment, a two-phase min-max optimization, and VLA-specific attention and semantic losses.
This paper proposes a communication-constrained multi-agent reinforcement learning framework that utilizes a generalized model and dual mutual information estimator to distinguish between lossy and lossless messages, thereby quantifying their impact on global rewards to enhance cooperative policy learning in complex, dynamic environments.
This paper introduces ELERAG, an enhanced Retrieval-Augmented Generation system that integrates Wikidata-based Entity Linking and a hybrid re-ranking strategy to significantly improve factual accuracy in Italian educational question-answering, particularly outperforming standard methods in domain-specific contexts while demonstrating the importance of domain-adapted strategies.
This paper introduces EMFusion, a conditional multivariate diffusion-based framework that leverages a residual U-Net with cross-attention and imputation-based sampling to provide accurate, uncertainty-quantified, frequency-selective electromagnetic field forecasts for wireless network planning, significantly outperforming existing baseline models.
This paper demonstrates that a single-epoch, domain-adapted fine-tuning of a 350M-parameter Small Language Model (OPT-350M) can significantly outperform larger models and existing baselines in tool-calling tasks, achieving a 77.55% pass rate on ToolBench and proving that targeted training can make generative AI more cost-effective and scalable for enterprise use.
This paper introduces SAGE, a novel Reinforcement Learning framework that enhances LLM-based agents' self-improvement capabilities by utilizing a skill library with sequential rollouts and skill-integrated rewards, achieving significantly higher goal completion rates and greater efficiency than existing methods on the AppWorld benchmark.
The paper proposes Manifold-Consistent Graph Indexing (MCGI), a geometry-aware, disk-resident indexing method that leverages Local Intrinsic Dimensionality to dynamically adapt search strategies, achieving significantly higher throughput and lower latency than state-of-the-art baselines on billion-scale datasets by resolving the Euclidean-Geodesic mismatch in high-dimensional spaces.