SHINE: A Scalable In-Context Hypernetwork for Mapping Context to LoRA in a Single Pass
Explorar
Noticias de IA
27413 elementos — filtrados, clasificados y sin duplicados
Segmentation before Answering: Pixel Grounding for MLLM Visual Reasoning
Beyond the Leaderboard: A Synthesis of Tool-Use, Planning, and Reasoning Failures in Larg…
ArtisanCAD: An Industrial-Level CAD Agent with Expert-Grounded Knowledge Distillation
Two Sides of the Same Coin: Learning the Backdoor to Remove the Backdoor
Association Restoration Test: Revealing Restorable Shortcuts after Unlearning
SpanUQ: Span-Level Uncertainty Quantification for Large Language Model Generation
Akashic: A Low-Overhead LLM Inference Service with MemAttention
Anthropic J-space research 🧠, Apple + Broadcom ⚡, continual agent learning 🤖
Small AI Models Gain Traction In places with unreliable networks
Teaching models to forget: Selective unlearning with Amazon Nova
The yes-no bias of large language models reflects answer order and wording, not shifts in…
Statistical Adversaries: Natural Backdoor-like Features in Vision Datasets
From Fixed to Free Cameras: Calibration-Free View-Robust Vision-Language-Action Model
Weak-to-Strong Generalization via Direct On-Policy Distillation
LLM-as-a-Verifier: A General-Purpose Verification Framework
Deform360: A Massive Multi-view Visuotactile Dataset for Deformable World Models
InFlux++: Real and Synthetic Data for Estimating Dynamic Camera Intrinsics
Fitted Occupancy-Ratio Evaluation without Bellman Completeness
SPEARBench: A Benchmark for Naturalness Evaluation in Streaming Speech-to-Speech Language…
REDDIT: Correcting Model-Generated Timestamp Drift in ASR without Forgetting via Replay-B…
Graph Sparse Sampling: Breaking the Curse of the Horizon in Continuous MDP Planning
ReCal3R: Reliability-Calibrated Learning Rates for Streaming 3D Reconstruction
TREK: Distill to Explore, Reinforce to Refine
How Much is Left? LLMs Linearly Encode Their Remaining Output Length
Deep Learning for Semen Analysis in Male Infertility: Computer Vision, Multimodal Fusion,…
Biologically Informed Deep Neural Networks for Multi-Omic Integration, Pathway Activity I…
Learning Only What Valid Adapters Can Express: Subspace-Constrained Adaptation Against Fi…
Wavelet Scattering Transform for Interpretable Schizophrenia Biomarker Discovery and Clas…
Erasing Without Collateral Damage: Precise Concept Removal in Diffusion Models