KI-Forschungspapiere Täglich

Meteor: Mamba-based Traversal of Rationale for Large Language and Vision Models

ConvLLaVA: Hierarchical Backbones as Visual Encoder for Large Multimodal Models

Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization

Aya 23: Open Weight Releases to Further Multilingual Progress

Stacking Your Transformers: A Closer Look at Model Growth for Efficient LLM Pre-Training

The Road Less Scheduled

AutoCoder: Enhancing Code Large Language Model with AIEV-Instruct

CraftsMan: High-fidelity Mesh Generation with 3D Native Generation and Interactive Geometry Refiner

Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach

iVideoGPT: Interactive VideoGPTs are Scalable World Models

Denoising LM: Pushing the Limits of Error Correction Models for Speech Recognition

Data Mixing Made Efficient: A Bivariate Scaling Law for Language Model Pretraining

HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting