HuggingFace Blog
101 posts indexed since 2026
Post Details
| Title | Author | Published | Words | HN Pts |
|---|---|---|---|---|
| Red Teaming with RL: Exploiting Tinker API for Harmful RL on 235B … | weitaofeng | 2026-01-01 | 1,778 | -- |
| The Engineering Handbook for GRPO + LoRA with Verl: Training Qwen2.5 on … | Yağız Çalık | 2026-01-02 | 5,072 | -- |
| Introducing Falcon-H1-Arabic: Pushing the Boundaries of Arabic Language AI with Hybrid Architecture | Basma Boussaha, Mohammed Alyafeai, Ahmed Alzubaidi, Leen AlQadi, Shaikha Alsuwaidi, Omar saif alkaabi, Hamza Alobeidli, and Hakim Hacid | 2026-01-05 | 1,838 | -- |
| TFLOPS Gap: Why FP4 MoE Kernel Engineering Matters on Blackwell | Konstantin | 2026-01-05 | 3,309 | -- |
| Introducing Falcon H1R 7B | Iheb Chaabane, Puneesh Khanna, Suhail M Shah, Slim Frikha, Shi Hu, Abdalgader Abubaker, Reda alami, Mike Lubinets, Mohamed El Amine Seddik, and Hakim Hacid | 2026-01-05 | 1,332 | -- |
| Building Autonomous Vehicles That Reason with the NVIDIA Alpamayo Open Ecosystem | Marco Pavone | 2026-01-05 | 893 | -- |
| Understanding Low-Rank Adaptation (LoRA): A Revolution in Fine-Tuning Large Language Models | Ashish Chadha | 2026-01-03 | 2,023 | -- |
| NVIDIA Cosmos Reason 2 Brings Advanced Reasoning To Physical AI | Tsung-Yi Lin and Debraj Sinha | 2026-01-05 | 1,037 | -- |
| Scaling Real-Time Voice Agents with Cache-Aware Streaming ASR | Kunal Dhawan, Adi- margolin, Gordana Neskovic, Maryam Motamedi, and Yasmina Benkhoui | 2026-01-05 | 1,860 | -- |
| NVIDIA brings agents to life with DGX Spark and Reachy Mini | Jeff Boudier, Nader Khalil, and Alec Fong | 2026-01-05 | 2,128 | -- |
| M2.1: Multilingual and Multi-Task Coding with Strong Generalization | MiniMax | 2026-01-05 | 2,306 | -- |
| Generalist Robot Policy Evaluation in Simulation with NVIDIA Isaac Lab-Arena and LeRobot | Raffaello Bonghi, lior ben horin, Kartik S, and Kalyan Vadrevu | 2026-01-05 | 1,038 | -- |
| Small Yet Mighty: Improve Accuracy In Multimodal Search and Visual Document Retrieval … | Ronay Ak, Gabriel de Souza Pereira Moreira, and Bo Liu | 2026-01-06 | 1,492 | -- |
| OpenMed: Six Months of Open-Source Medical AI and the Road Ahead | Maziyar Panahi | 2026-01-06 | 2,424 | -- |
| Why We Built VIBE Bench: Rethinking Evaluation for Real Workloads | MiniMax | 2026-01-06 | 736 | -- |
| Diversity Vs Density: A data strategy comparison for fine-tuning VLMs | Akhil Theerthala | 2026-01-06 | 2,301 | -- |
| 🥃 Distilling Tiny Embeddings | David Mezzetti | 2026-01-10 | 1,082 | -- |
| Introducing OptiMind, a research model designed for optimization | Anson Ho, Sirui Li, and Ishai Menache | 2026-01-15 | 395 | -- |
| How We Built a Semantic Highlight Model To Save Token Cost for … | Cheney Zhang and Jiang Chen | 2026-01-15 | 2,344 | -- |
| Proof of Time: A Benchmark for Evaluating Scientific Idea Judgments | Bingyang Ye and Shan Chen | 2026-01-13 | 2,717 | -- |
| Open Responses: What you need to know | shaun smith, ben burtenshaw, merve, and Pedro Cuenca | 2026-01-15 | 1,344 | -- |
| Beyond Brute Force: Why LoongFlow is the “Thinking” Evolution of OpenEvolve | Xunan Dai | 2026-01-16 | 1,108 | -- |
| SmolLM-Smashed: Tiny Giants, Optimized for Speed | David Berenstein | 2026-01-13 | 982 | -- |
| VLM-OCR Recipes on GPU Infrastructure | Florent Gbelidji | 2026-01-15 | 2,281 | -- |
| Reviewer Two (but it's an OpenEnv) | Chris von Csefalvay | 2026-01-13 | 1,653 | -- |
| Scaling OpenEnv: From Free Usage to Thousands of Concurrent Environments | ben burtenshaw | 2026-01-20 | 1,158 | -- |
| LightOnOCR-2-1B: a lightweight high-performance end-to-end OCR model family | Said Taghadouini, Adrien Cavaillès, and Baptiste Aubertin | 2026-01-19 | 934 | -- |
| Differential Transformer V2 | Li Dong | 2026-01-20 | 3,136 | -- |
| 🪄 Interpreto: A Unified Toolkit for Interpretability of Transformer Models | Fanny Jourdan and Antonin Poché | 2026-01-20 | 2,112 | -- |
| New in llama.cpp: Anthropic Messages API | Xuan-Son Nguyen and Victor Mustar | 2026-01-19 | 541 | -- |
| One Year Since the “DeepSeek Moment” | Adina Yakefu and Irene Solaiman | 2026-01-20 | 1,617 | -- |
| Optimizing GLM4-MoE for Production: 65% Faster TTFT with SGLang | Novita AI | 2026-01-22 | 1,047 | -- |
| Security, Governance and Performance for Dell On-Prem AI Builders | Balachandran Rajendran, Juan Julián, Alvaro Bartolome, Enrique Hernández Calabrés, Simon Pagezy, and Jeff Boudier | 2026-01-21 | 1,064 | -- |
| RexRerankers: SOTA Rankers for Product Discovery and AI Assistants | Rahul Bajaj, Anuj Garg, and Jaya Nupur | 2026-01-24 | 3,704 | -- |
| Challenges of Synthetic Dataset Generation | Rishiraj Acharya | 2026-01-21 | 942 | -- |
| Reverse Engineering a $500M Mystery: From HashHop to Memory-Augmented Language Models | Asankhaya Sharma | 2026-01-23 | 1,825 | -- |
| AssetOpsBench: Bridging the Gap Between AI Agent Benchmarks and Industrial Reality | Dhaval Patel, James Rayfield, Saumya Ahuja, Chathurangi Shyalika, Shuxin Lin, and Zhou | 2026-01-21 | 1,505 | -- |
| “DeepSeek R1 时刻” 一周年 | vansin | 2026-01-20 | 315 | -- |
| Benchmark Smarter: Tailor Your Model Evaluation Suite with EvalScope | kelseye.xh | 2026-01-22 | 1,973 | -- |
| Waypoint-1: Real-time Interactive Video Diffusion from Overworld | Andrew Lapp, Louis Castricato, Scott Fox, Shahbuland Matiana, and David Rossi | 2026-01-20 | 853 | -- |
| Why Your AI Strategy Needs Hugging Face Storage | Adrian Lepers | 2026-01-26 | 1,008 | -- |
| NVIDIA Earth-2 Open Models Span the Whole Weather Stack | Mike Pritchard, Jaideep Pathak, Jean Kossaifi, and Aayush Gupta | 2026-01-26 | 736 | -- |
| Alyah ⭐️: Toward Robust Evaluation of Emirati Dialect Capabilities in Arabic LLMs | Omar saif alkaabi, Ahmed Alzubaidi, Hamza Alobeidli, Shaikha Alsuwaidi, Mohammed Alyafeai, Leen AlQadi, Basma Boussaha, and Hakim Hacid | 2026-01-27 | 1,585 | -- |
| Unlocking Agentic RL Training for GPT-OSS: A Practical Retrospective | Jason Zhu, Hejian Sang, Arup De, Rohit Jain, and Yanning Chen | 2026-01-27 | 4,160 | -- |
| Friends and Grandmothers in Silico | Itay Yona | 2026-01-24 | 4,089 | -- |
| Architectural Choices in China's Open-Source AI Ecosystem: Building Beyond DeepSeek | Adina Yakefu and Irene Solaiman | 2026-01-27 | 1,324 | -- |
| Nemotron-Personas-Brazil: Co-Designed Data for Sovereign AI | Andre Manoel, Yev Meyer, Shyamala Prayaga, Will Jennings, and bardiya sadeghi | 2026-01-28 | 903 | -- |
| The Great Classification Showdown: OSS vs BERT on Consumer Hardware | Ben Toussaint | 2026-01-26 | 1,938 | -- |
| We got Claude to teach open models how to write CUDA kernels! | ben burtenshaw, shaun smith, merve, and Pedro Cuenca | 2026-01-28 | 2,350 | -- |
| Slashing torch.compile Warmup & LoRA Swapping Times with Pruna | John Rachwan, Johanna Sommer, Bertrand Charpentier, and Sara Han Díaz | 2026-01-28 | 1,513 | -- |
| Nemotron-Personas-Singapore: Co-Designed Data for Sovereign AI | Will Jennings, Dane Corneil, Yev Meyer, Verdi March, Shyamala Prayaga, and bardiya sadeghi | 2026-01-27 | 1,041 | -- |
| TruthTensor: LLM Evalution in Prediction Markets Under Drift and Market Baseline | Elena Pashkova, shirin Shahabi, Hudson, and Ronald Chan | 2026-01-29 | 1,631 | -- |
| Performant local mixture-of-experts CPU inference with GPU acceleration in llama.cpp | Doctor Shotgun and Geechan | 2026-01-30 | 2,508 | -- |
| Introducing NVIDIA Cosmos Policy for Advanced Robot Control | Pranjali Joshi, Tsung-Yi Lin, Jinwei Gu, and Prachi Mishra | 2026-01-29 | 1,333 | -- |
| Introducing Daggr: Chain apps programmatically, inspect visually | merve, yuvraj sharma, Abubakar Abid, hysts, and Pedro Cuenca | 2026-01-29 | 1,559 | -- |
| Fine-Tuning FunctionGemma on TPU to Create a Virtual Fitness Coach in 10 … | Alvaro Moran | 2026-02-02 | 2,906 | -- |
| Announcing ReasoningLens — Visualizing and Diagnosing LLM Reasoning at a Glance | Jun Zhang, Jason Zheng, Boxi Cao, and ReasoningLens | 2026-02-03 | 693 | -- |
| Training Design for Text-to-Image Models: Lessons from Ablations | David Bertoin, Roman Frigg, and Jon Almazán | 2026-02-03 | 7,420 | -- |
| The Future of the Global Open-Source AI Ecosystem: From DeepSeek to AI+ | Adina Yakefu and Irene Solaiman | 2026-02-03 | 1,602 | -- |
| H Company's new Holo2 model takes the lead in UI Localization | Ramzi De Coster, Hamza Benchekroun, and Aurélien Lac | 2026-02-03 | 214 | -- |
| Nemotron ColEmbed V2: Raising the Bar for Multimodal Retrieval with ViDoRe V3’s … | Ronay Ak and Gabriel de Souza Pereira Moreira | 2026-02-04 | 1,048 | -- |
| Nvidia Agentic Smart Router on Dell Enterprise Hub : Deepdive on Architecture,Design … | Khushboo Rathi and Balachandran Rajendran | 2026-02-03 | 995 | -- |
| CRAFT: Continuous Reasoning and Agentic Feedback Tuning | Valentin, Denis Timonin, Alexandr, and Alexey | 2026-02-05 | 813 | -- |
| Introducing SyGra Studio | Surajit Dasgupta, Bidyapati Pradhan, Amit Kumar Saha, Vipul Mittal, and Sriram Puttagunta | 2026-02-05 | 747 | -- |
| 🚀 SyGra V2.0.0 | Sriram Puttagunta, Surajit Dasgupta, Bidyapati Pradhan, Amit Kumar Saha, and Vipul Mittal | 2026-02-05 | 724 | -- |
| From Golden Gate Bridge to Broken JSON: Why Anthropic's SAE Steering Fails … | Maziyar Panahi | 2026-02-07 | 5,766 | -- |
| Transformers.js v4 Preview: Now Available on NPM! | Joshua and Nico Martin | 2026-02-09 | 1,185 | -- |
| Training Qwen3 VL to label bbox : synthetic data, environment and training … | Ulrick BLE | 2026-02-09 | 2,544 | -- |
| 🚀 DTS: A Candidate for the Best Parallel Reasoning in LLMs | Guanchu | 2026-02-11 | 616 | -- |
| Building a Mood-Based Movie Recommendation Engine with Voyage-4-nano, Hugging Face, and MongoDB … | Arkadiusz Borucki | 2026-02-08 | 3,315 | -- |
| Enabling Large Scale RLHF of GPTOSS with Megatron backend in VeRL | LEI WANG | 2026-02-10 | 5,934 | -- |
| OpenEnv in Practice: Evaluating Tool-Using Agents in Real-World Environments | Christian Washington, Ankit Jasuja, Santosh Sah, Lewis Tunstall, and ben burtenshaw | 2026-02-12 | 1,656 | -- |
| LateOn-Code & ColGrep: LightOn unveils state-of-the-art code retrieval models and code search … | Antoine Chaffin and Raphael | 2026-02-12 | 4,993 | -- |
| Forge: Scalable Agent RL Framework and Algorithm | MiniMax, Hyn, zhi zhang, Jiayuan Song, Da Chen, xkc, Yaoyao, kennyKK, and zpysky1125 | 2026-02-13 | 3,387 | -- |
| How to Use Multiple GPUs in Hugging Face Transformers: Device Map vs … | Aritra Roy Gosthipaty | 2026-02-12 | 606 | -- |
| Custom Kernels for All from Codex and Claude | ben burtenshaw, Sayak Paul, Aritra Roy Gosthipaty, and shaun smith | 2026-02-13 | 1,792 | -- |
| What superpower does Kimi-K2.5 bring to the table? | Leco Li | 2026-02-13 | 1,154 | -- |
| The Chinese GLM-5 Model Now Ranks #2 in Arabic Language Performance | Karim Ouda | 2026-02-16 | 322 | -- |
| Compute and Competition in AI: Different FlOPs for Different Folks | Yacine Jernite and Sasha Luccioni | 2026-02-12 | 1,917 | -- |
| How to Build a Benchmark with a Private Test Set on Hugging … | Georgia Channing | 2026-02-16 | 1,775 | -- |
| Qwen3.5: Nobody Agrees on Attention Anymore | Maxime Labonne | 2026-02-17 | 1,192 | -- |
| NVIDIA Nemotron 2 Nano 9B Japanese: 日本のソブリンAIを支える最先端小規模言語モデル | Atsunori Fujita, Kotaro Yamamoto, Masaya Ogushi, Vincent Gong, Ameya Sunil Mahabaleshwarkar, and Yoshi Suhara | 2026-02-17 | 297 | -- |
| DenseR: Dense Rewards For Free in LLM Reasoning | Hritik Bansal | 2026-02-18 | 3,977 | -- |
| De-mystifying Multimodal Learning: Enabiling Vision in Language Models | Matteo Nulli | 2026-02-17 | 2,797 | -- |
| One-Shot Any Web App with Gradio's gr.HTML | yuvraj sharma, hysts, and Freddy Boulton | 2026-02-18 | 829 | -- |
| IBM and UC Berkeley Diagnose Why Enterprise Agents Fail Using IT-Bench and … | Ayhan Sebin, Rohan Arora, and Saurabh Jha | 2026-02-18 | 2,253 | -- |
| Did GPT 5.2 make a breakthrough discovery in theoretical physics? | David Louapre | 2026-02-19 | 4,541 | -- |
| ColBERT-Zero: To Pre-train Or Not To Pre-train ColBERT models? | Antoine Chaffin, Luca Arnaboldi, Amélie Chatelain, and Florent Krzakala | 2026-02-19 | 2,306 | -- |
| 「データ不足」の壁を越える：合成ペルソナが日本のAI開発を加速 | Atsunori Fujita, Masaya Ogushi, Will Jennings, Yev Meyer, Kotaro Yamamoto, Yoshi Suhara, Vincent Gong, and Dane Corneil | 2026-02-19 | 280 | -- |
| I Let a Lobster Run My Jetson: What OpenClaw Taught Me About … | Andres Marafioti | 2026-02-19 | 1,509 | -- |
| Train AI models with Unsloth and Hugging Face Jobs for FREE | ben burtenshaw, Daniel (Unsloth), Michael Han, Maxime Labonne, Daniel van Strien, and shaun smith | 2026-02-20 | 944 | -- |
| GGML and llama.cpp join HF to ensure the long-term progress of Local … | Georgi Gerganov, Xuan-Son Nguyen, Aleksander Grygier, Lysandre, Victor Mustar, and Julien Chaumond | 2026-02-20 | 936 | -- |
| Introducing Legal RAG Bench | Umar Butler and Abdur-Rahman Butler | 2026-02-20 | 3,235 | -- |
| FINAL Bench: The Real Bottleneck to AGI Is Self-Correction | VIDRAFT_LAB | 2026-02-21 | 1,146 | -- |
| How We Learned to Talk to Machines | Tyler Williams | 2026-02-20 | 1,156 | -- |
| Kimi K2.5: Still Worth It After Two Weeks? | Maxime Labonne | 2026-02-23 | 1,448 | -- |
| Do Bubbles Form When Tens of Thousands of AIs Simulate Capitalism? | VIDRAFT_LAB | 2026-02-24 | 2,770 | -- |
| Follow the White Rabbit: Using Embeddings So You Never Get Lost in … | David Corvoysier | 2026-02-23 | 1,420 | -- |
| MAEB: Evaluating Audio Embeddings at Scale | Adnan El Assadi, Solomatin Roman, Kenneth C. Enevoldsen, and Isaac Chung | 2026-02-24 | 1,349 | -- |
| A Guide to Reinforcement Learning Post-Training for LLMs: PPO, DPO, GRPO, and … | Karina Zadorozhny | 2026-01-19 | 7,738 | -- |
| Deploying Open Source Vision Language Models (VLM) on Jetson | Mitesh Patel, Johnny Nuñez Cano, and Raymond Lo | 2026-02-24 | 1,591 | -- |