Roman Potemin

potemin

·

poteminr

AI & ML interests

nlp, ner, multi-agent systems, code agents

Recent Activity

updated a collection about 15 hours ago

Production models

updated a collection about 15 hours ago

Production models

updated a collection about 15 hours ago

Production models

View all activity

Organizations

upvoted a collection 4 months ago

DINOv3

DINOv3: foundation models producing excellent dense features, outperforming SotA w/o fine-tuning - https://arxiv.org/abs/2508.10104 • 15 items • Updated Mar 10 • 752

upvoted 2 collections 6 months ago

FastVLM

Efficient Vision Encoding for Vision Language Models • 8 items • Updated Mar 2 • 114

MobileCLIP2

MobileCLIP2: Mobile-friendly image-text models with SOTA zero-shot capabilities trained on DFNDR-2B • 30 items • Updated Apr 23 • 64

upvoted an article 6 months ago

Article

SigLIP 2: A better multilingual vision language encoder

+1

ariG23498, merve, qubvel-hf

•

Feb 21, 2025

• 226

upvoted a paper 6 months ago

SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features

Paper • 2502.14786 • Published Feb 20, 2025 • 169

upvoted a collection 6 months ago

SigLIP2

36 items • Updated Jul 21 • 123

upvoted a collection over 1 year ago

ModernBERT

Bringing BERT into modernity via both architecture changes and scaling • 3 items • Updated Dec 19, 2024 • 161

upvoted a collection over 2 years ago

AQLM

AQLM quantized LLMs • 21 items • Updated Feb 18 • 46