Back to Blog

Paper library

Reading, questioning and connecting ideas in machine learning. Notes from our ongoing exploration.

All notes

344 notes
From the articleOriginal paper
arXiv 2026ICDAR Workshop 2026Model evaluation

Low CER

Right Letters, Wrong Person

When Low CER is Not Enough: An Analysis of Hallucinations in Vision-Language OCR Systems on Historical Uruguayan Documents

Note published
From the articleOriginal paper
arXiv 2026ICML 2026Classic CNNs

WBMM

Bigger Windows, Higher Speed

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

Note published
From the articleOriginal paper
arXiv 2026Object detection

YOLO26

Keeping Both Heads

Ultralytics YOLO26: Unified Real-Time End-to-End Vision Models

Note published
From the articleOriginal paper
arXiv 2026Findings of EMNLP 2026Model evaluation

FaithC4

The Overeager Proofreader

Do VLMs Read or Rewrite? On Transcription Faithfulness in Vision-Language Models

Note published
From the articleOriginal paper
arXiv 2026ICLR 2026State space models

Mamba-3

Inference First

Mamba-3: Improved Sequence Modeling using State Space Principles

Note published
From the articleOriginal paper
arXiv 2026CVPR 2026Classic CNNs

FCDM

Must It Be a Transformer?

Reviving ConvNeXt for Efficient Convolutional Diffusion Models

Note published
Showing 12 of 344