KI-Forschungspapiere Täglich

Jamba: A Hybrid Transformer-Mamba Language Model

Gecko: Versatile Text Embeddings Distilled from Large Language Models

Transformer-Lite: High-efficiency Deployment of Large Language Models on Mobile Phone GPUs

ReALM: Reference Resolution As Language Modeling

InstantSplat: Unbounded Sparse-view Pose-free Gaussian Splatting in 40 Seconds

Unsolvable Problem Detection: Evaluating Trustworthiness of Vision Language Models

Localizing Paragraph Memorization in Language Models

DiJiang: Efficient Large Language Models through Compact Kernelization

MambaMixer: Efficient Selective State Space Models with Dual Token and Channel Selection

Snap-it, Tap-it, Splat-it: Tactile-Informed 3D Gaussian Splatting for Reconstructing Challenging Surfaces