Blog, Articles, and discussions

Community Articles

LightOnOCR-1B: The Case for End-to-End and Efficient Domain-Specific Vision-Language Models for OCR

and 2 others •

Introducing MTEB v2: Evaluation of embedding and retrieval systems for more than just text

and 2 others •

Australian-made LLM beats OpenAI and Google at legal retrieval

and 2 others •

Promoter-GPT: Writing DNA Instructions with Language Models

GSMA Open-Telco LLM Benchmarks 2.0: The first dedicated LLM Evaluation for Telecoms

and 15 others •

How I Built Lightning-Fast Vector Search for Legal Documents

Art of Focus: Page-Aware Sparse Attention and Ling 2.0’s Quest for Efficient Context Length Scaling

and 19 others •

Scaling Test-Time Compute to Achieve Gold Medal at IOI 2025 with Open-Weight Models

and 3 others •

Llama‑Embed‑Nemotron‑8B Text Embedding Model Ranks First on Multilingual MTEB Leaderboard

and 4 others •

AI for Food Allergies

and 3 others •

Announcing Hugging Face Fundamentals: A New Learning Track on DataCamp

Nemotron’s Open Secret: Accelerating AI Development with Open Models, Data, and Recipes

and 1 other •

Uncensor any LLM with abliteration

Merge Large Language Models with mergekit

KV Caching Explained: Optimizing Transformer Inference Efficiency

Model statistics of the 50 most downloaded entities on Hugging Face

There is no such thing as a tokenizer-free lunch

about 1 month ago

How Financial News Can Be Used to Train Good Financial Models

Vision Tokens vs Text Tokens: Understanding the 10× Compression

Post-Training Isaac GR00T N1.5 for LeRobot SO-101 Arm

and 5 others •

nlpcommunityresearch

Google releases Gemma 2 2B, ShieldGemma and Gemma Scope

communityevaluationsynthetic-data

LAVE: Zero-shot VQA Evaluation on Docmatix with LLMs - Do We Still Need Fine-Tuning?

nlpcommunityresearch

Llama 3.1 - 405B, 70B & 8B with multilinguality and long context

+4

communitydatasetssynthetic-data

Docmatix - a huge dataset for Document Visual Question Answering

llmnlpsynthetic-data

SmolLM - blazingly fast and remarkably powerful

ai4mathnlpcommunity

How NuminaMath Won the 1st AIMO Progress Prize

+4

agentssmolagentsnlp

Our Transformers Code Agent beats the GAIA benchmark!

nlpcommunityresearch

Welcome Gemma 2 - Google's new open LLM

+2

collaborationcommunityopen-source

Fine-tuning Florence-2 - Microsoft's Cutting-edge Vision Language Models

leaderboardresearchcollaboration

BigCodeBench: Benchmarking Large Language Models on Solving Practical and Challenging Programming Tasks

+5

open-sourceguideresearch

From DeepSpeed to FSDP and Back Again with Hugging Face Accelerate

Putting RL back in RLHF

leaderboardresearchcollaboration

Launching the Artificial Analysis Text to Image Leaderboard & Arena

nlpcommunityresearch

Falcon 2: An 11B parameter pretrained language model and VLM, trained on over 5000B tokens tokens and 11 languages

+5

Community Articles

LightOnOCR-1B: The Case for End-to-End and Efficient Domain-Specific Vision-Language Models for OCR

and 2 others •

Introducing MTEB v2: Evaluation of embedding and retrieval systems for more than just text

and 2 others •

Australian-made LLM beats OpenAI and Google at legal retrieval

and 2 others •

Promoter-GPT: Writing DNA Instructions with Language Models

GSMA Open-Telco LLM Benchmarks 2.0: The first dedicated LLM Evaluation for Telecoms

and 15 others •

How I Built Lightning-Fast Vector Search for Legal Documents

Art of Focus: Page-Aware Sparse Attention and Ling 2.0’s Quest for Efficient Context Length Scaling

and 19 others •

Scaling Test-Time Compute to Achieve Gold Medal at IOI 2025 with Open-Weight Models

and 3 others •

Llama‑Embed‑Nemotron‑8B Text Embedding Model Ranks First on Multilingual MTEB Leaderboard

and 4 others •

AI for Food Allergies

and 3 others •

Announcing Hugging Face Fundamentals: A New Learning Track on DataCamp

Nemotron’s Open Secret: Accelerating AI Development with Open Models, Data, and Recipes

and 1 other •

Uncensor any LLM with abliteration

Merge Large Language Models with mergekit

KV Caching Explained: Optimizing Transformer Inference Efficiency

Model statistics of the 50 most downloaded entities on Hugging Face

There is no such thing as a tokenizer-free lunch

about 1 month ago

How Financial News Can Be Used to Train Good Financial Models

Vision Tokens vs Text Tokens: Understanding the 10× Compression

Post-Training Isaac GR00T N1.5 for LeRobot SO-101 Arm

and 5 others •

View all