GenDA is a generative data assimilation framework that uses graph-based diffusion models to reconstruct high-resolution urban wind fields from sparse sensor data.
A new curvature-aware optimization framework uses minimum description length principles to guide layer-adaptive capacity allocation and pruning in large language models.
The proposed PolyUQuest framework improves web-based retrieval-augmented generation by utilizing heterogeneous graphs to preserve the structural and semantic layout of HTML pages.
The proposed APIVOT system combines linguistic and visual reasoning to help robots plan and execute complex, long-horizon tasks while respecting physical constraints.
SolarChain-Eval is a new physics-constrained benchmark designed to evaluate the performance and trustworthiness of autonomous AI agents in decentralized energy markets.
The authors propose expanding the agentic inequality framework to include interaction-level architectural differences, highlighting how varying context access affects user experiences with AI agents.
A new face-swapping pipeline protects pedestrian privacy in autonomous vehicle training datasets while preserving critical behavioral features needed for trajectory prediction.
A new multi-agent framework utilizes cognitive appraisal theory to enable dynamic emotional evolution in persona-based dialogue systems.
A new industrial dataset covering machinery usage and failures in Nigeria has been released to support quantitative analysis and language model training for African economies.
This paper introduces a method that uses large language models as judges to align metrics for personality recognition without relying on rigid psychological theories.
This paper provides a structured analysis of the evaluation methodologies and design paradigms that have driven progress in deep reinforcement learning over the past decade.
TOPO-Bench is an open-source framework designed to standardize the evaluation of topological mapping systems and quantify the impact of perceptual aliasing.
The SHARP framework is proposed to help streaming sequence models learn long-range, non-stationary temporal patterns without needing to revisit past observations.
Researchers adapted the Joint Embedding Predictive Architecture to analyze compact network security fingerprints for threat detection.
Researchers developed Safe Flow Q-Learning, an offline reinforcement learning method that integrates a reachability-inspired safety value function to maintain strict safety constraints.
Researchers developed a method to optimize token generation ordering in masked diffusion models to improve performance in multimodal understanding and image synthesis.
As AI inference costs decline rapidly, researchers are exploring how data systems must evolve to support, manage, and be driven by autonomous AI agents.
Inspired by cognitive neuroscience, this paper proposes a multi-view hypergraph learning method to improve the detection of online financial fraud.
The Octree Residual Network enables real-time, differentiable Euclidean signed distance function reconstruction from point cloud data for robotic autonomy.
Researchers have modeled psychological disorder phenotypes in reinforcement learning agents by manipulating cognitive appraisal signals within an appraisal-guided optimization framework.
Researchers have developed EasyLens, a training-free, plug-and-play tool designed to enhance the sensitivity of medical vision-language models to subtle clinical lesions.
SimRPD is a new framework designed to optimize proactive recruitment dialogue agents by evaluating and selecting high-quality training data using a simulator.
ArtMine is a framework designed to analyze and formalize the iterative decision-making processes of creating art rather than just modeling the final visual output.
FedOPAL introduces an analytical visual prompt tuning method to reduce communication overhead and server-side computation in one-shot federated learning.