π€ Just published: "Consent by Design" - exploring how we're building better consent mechanisms across the HF ecosystem!
Our research shows open AI development enables: - Community-driven ethical standards - Transparent accountability - Context-specific implementations - Privacy as core infrastructure
Check out our Space Privacy Analyzer tool that automatically generates privacy summaries of applications!
Effective consent isn't about perfect policies; it's about architectures that empower users while enabling innovation. π
Hacked my presentation building with inference providers, Cohere command a, and sheer simplicity. Use this script if youβre burning too much time on presentations:
This is what it does: - uses command a to generates slides and speaker notes based on some material. - it renders the material in remark open format and imports all images, tables, etc - you can then review the slides as markdown and iterate - export to either pdf or pptx using backslide
π Next steps are: add text to speech for the audio and generate a video. This should make Hugging Face educational content scale to a billion AI Learners.
reacted to yjernite's
post with π₯about 17 hours ago
Today in Privacy & AI Tooling - introducing a nifty new tool to examine where data goes in open-source apps on π€
HF Spaces have tons (100Ks!) of cool demos leveraging or examining AI systems - and because most of them are OSS we can see exactly how they handle user data ππ
That requires actually reading the code though, which isn't always easy or quick! Good news: code LMs have gotten pretty good at automatic review, so we can offload some of the work - here I'm using Qwen/Qwen2.5-Coder-32B-Instruct to generate reports and it works pretty OK π
The app works in three stages: 1. Download all code files 2. Use the Code LM to generate a detailed report pointing to code where data is transferred/(AI-)processed (screen 1) 3. Summarize the app's main functionality and data journeys (screen 2) 4. Build a Privacy TLDR with those inputs
It comes with a bunch of pre-reviewed apps/Spaces, great to see how many process data locally or through (private) HF endpoints π€
You can now bill your inference costs from all our inference partners (together, fireworks, fal, sambanova, cerebras, hyperbolic,...) to your Hugging Face organization.
Useful to drive more company-wide usage of AI without the billing headaches!
reacted to gavinkhung's
post with π€about 17 hours ago
Collection of 536,231 question-answer pairs featuring:
- Human-posed questions and machine-generated responses for SFT - Bilingual content in Russian and English with linked IDs - Derived from 739k+ real user queries, primarily educational topics - Includes unique IDs and machine-generated category labels
This dataset provides a resource for supervised fine-tuning (SFT) of large language models, cross-lingual research, and understanding model responses to diverse user prompts. Released to the public domain under CC0 1.0 license.
reacted to JLouisBiz's
post with πabout 17 hours ago
This is short demonstration of large language model integration into a user's workflow. This is helping to quickly save or capture whatever you have copied to your clipboard. It goes into the database. In your case, it could go to the file. It could be published quickly. You could make a one-click page or one-click document. Eventually, it becomes immediately a note for later use.
We are excited to announce the release of our paper, "Cobra: Efficient Line Art COlorization with BRoAder References," along with the official code! Cobra is a novel efficient long-context fine-grained ID preservation framework for line art colorization, achieving high precision, efficiency, and flexible usability for comic colorization. By effectively integrating extensive contextual references, it transforms black-and-white line art into vibrant illustrations.
Introducing BioClinicalBERT-Triage: A Medical Triage Classification Model I'm excited to share my latest project: a fine-tuned model for medical triage classification! What is BioClinicalBERT-Triage? BioClinicalBERT-Triage is a specialized model that classifies patient-reported symptoms into appropriate triage categories. Built on the foundation of emilyalsentzer/Bio_ClinicalBERT, this model helps healthcare providers prioritize patient care by analyzing symptom descriptions and medical history. Why I Built This As healthcare systems face increasing demands, efficient triage becomes crucial. This model aims to support healthcare professionals in quickly assessing the urgency of medical situations, particularly in telehealth and high-volume settings. Model Performance The model was trained on 42,513 medical symptom descriptions, using an 80:20 train/test split. After 3 epochs of training, the model achieved:
Final training loss: 0.3246 Processing speed: 13.99 samples/second
The loss steadily decreased throughout training:
Early training (epoch 0.24): 0.5796 Mid-training (epoch 1.65): 0.4308 Final (epoch 2.82): 0.3246 from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
Limitations & Ethical Considerations This model is designed to support, not replace, clinical decision-making. It should always be used under the supervision of qualified healthcare professionals. While it performs well on common presentations, it may be less accurate for rare conditions or unusual symptom descriptions. Try It Out I'd love to hear your feedback if you use this model in your projects! Check out the full model card here: VolodymyrPugachov/BioClinicalBERT-Triage #medical #healthcare #bert #nlp #triage #classification
reacted to thomwolf's
post with π€π₯3 days ago
If you've followed the progress of robotics in the past 18 months, you've likely noticed how robotics is increasingly becoming the next frontier that AI will unlock.
At Hugging Faceβin robotics and across all AI fieldsβwe believe in a future where AI and robots are open-source, transparent, and affordable; community-built and safe; hackable and fun. We've had so much mutual understanding and passion working with the Pollen Robotics team over the past year that we decided to join forces!
You can already find our open-source humanoid robot platform Reachy 2 on the Pollen website and the Pollen community and people here on the hub at pollen-robotics
We're so excited to build and share more open-source robots with the world in the coming months!
1 reply
Β·
reacted to thomwolf's
post with πβ€οΈ4 days ago
If you've followed the progress of robotics in the past 18 months, you've likely noticed how robotics is increasingly becoming the next frontier that AI will unlock.
At Hugging Faceβin robotics and across all AI fieldsβwe believe in a future where AI and robots are open-source, transparent, and affordable; community-built and safe; hackable and fun. We've had so much mutual understanding and passion working with the Pollen Robotics team over the past year that we decided to join forces!
You can already find our open-source humanoid robot platform Reachy 2 on the Pollen website and the Pollen community and people here on the hub at pollen-robotics
We're so excited to build and share more open-source robots with the world in the coming months!
1 reply
Β·
reacted to fdaudens's
post with ββ€οΈ10 days ago
I read the 456-page AI Index report so you don't have to (kidding). The wild part? While AI gets ridiculously more accessible, the power gap is actually widening:
1οΈβ£ The democratization of AI capabilities is accelerating rapidly: - The gap between open and closed models is basically closed: difference in benchmarks like MMLU and HumanEval shrunk to just 1.7% in 2024 - The cost to run GPT-3.5-level performance dropped 280x in 2 years - Model size is shrinking while maintaining performance - Phi-3-mini hitting 60%+ MMLU at fraction of parameters of early models like PaLM
2οΈβ£ But we're seeing concerning divides deepening: - Geographic: US private investment ($109B) dwarfs everyone else - 12x China's $9.3B - Research concentration: US and China dominate highly-cited papers (50 and 34 respectively in 2023), while next closest is only 7 - Gender: Major gaps in AI skill penetration rates - US shows 2.39 vs 1.71 male/female ratio
The tech is getting more accessible but the benefits aren't being distributed evenly. Worth thinking about as these tools become more central to the economy.
It's been a wild few days, and especially π€― to see every tensor file with a Xet logo next to it instead of LFS.
The attached graph shows requests per second to our content-addressed store (CAS) right as the release went live.
yellow = GETs; dashed line = launch time.
You can definitely tell when the community started downloading π
h/t to @rajatarya for the graph, the entire Xet crew to bring us to this point, and special shoutout to Rajat, @port8080, @brianronan , @seanses , and @znation who made sure the bytes kept flying all weekend β‘οΈ
AI agents are transforming how we interact with technology, but how sustainable are they? π
Design choices β like model size and structure β can massively impact energy use and cost. β‘π° The key takeaway: smaller, task-specific models can be far more efficient than large, general-purpose ones.
π Open-source models offer greater transparency, allowing us to track energy consumption and make more informed decisions on deployment. π± Open-source = more efficient, eco-friendly, and accountable AI.
- π§ Native Multimodality - Process text and images in a unified architecture - π Mixture-of-Experts - First Llama models using MoE for incredible efficiency - π Super Long Context - Up to 10M tokens - π Multilingual Power - Trained on 200 languages with 10x more multilingual tokens than Llama 3 (including over 100 languages with over 1 billion tokens each)
πΉ Llama 4 Scout - 17B active parameters (109B total) - 16 experts architecture - 10M context window - Fits on a single H100 GPU - Beats Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1
πΉ Llama 4 Maverick - 17B active parameters (400B total) - 128 experts architecture - It can fit perfectly on DGX H100(8x H100) - 1M context window - Outperforms GPT-4o and Gemini 2.0 Flash - ELO score of 1417 on LMArena currently second best model on arena
πΉ Llama 4 Behemoth (Coming Soon) - 288B active parameters (2T total) - 16 experts architecture - Teacher model for Scout and Maverick - Outperforms GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on STEM benchmarks
AutoGLM ζ²ζπ« FREE AI Agent released by ZhipuAI β¨ Think & Act simultaneously β¨ Based on a fully self-developed stack: GLM-4 for general, GLM-Z1 for inference, and GLM-Z1-Rumination for rumination β¨ Will openly share these models on April 14 π€―
