Accelerating LLM Inference via Vector Index Based Output Embeddings
Abstract page for arXiv paper 2608.27460: Accelerating LLM Inference via Vector Index Based Output Embeddings
Abstract page for arXiv paper 2608.27460: Accelerating LLM Inference via Vector Index Based Output Embeddings
Abstract page for arXiv paper 2608.27662: Unsupervised Continual Learning with Growing Self-Organizing Maps and Synthetic Replay
Abstract page for arXiv paper 2608.27634: Curvature-Aware Radius Shrinkage for Adaptive Nearest Neighbor Classification
Abstract page for arXiv paper 2608.27574: Self-Explainable Multi-Label Graph Neural Network for Correlated Evidence Attribution
Abstract page for arXiv paper 2608.27518: When Muon Meets Task Interference: A Spectral Perspective on Continual Learning and Model Merging
A theoretical preprint links resonance peaks, complex poles and residue phases, while showing where broad scalar states depart from elastic behavior.
A preprint predicts a semi-localized ground state in a disordered 1D model with exponential tail statistics and finite-size numerical support.
A preprint reports strong benchmark scores for TDFNet, a three-projection model for finding salient objects in 360-degree images.
An arXiv preprint tests geometry-constrained KANs, finding strong clean-data results but different performance under noise, small data and extrapolation.
An arXiv preprint tests primordial non-Gaussianity and early-Universe energy injection with ISW–tSZ data, finding model-dependent constraints.
A Bayesian framework maps uncertainty in discrete-choice probabilities into set-valued model answers and tests the approach in simulation.
A preprint links SkillShield to lower observed attack and malware-generation scores in coding-agent benchmarks, while live deployment remains untested.
Topics: Coding Agents
Entities: Coding Agents