23 articles
OpenAI develops a framework to evaluate risks of LLMs aiding biological threat creation. A study with biology experts and students shows GPT-4 provides at most mild uplift in biological threat creation accuracy, insufficient to be conclusive.
Hugging Face introduces a leaderboard focused on real-world enterprise use cases. It evaluates models on practical scenarios rather than traditional academic benchmarks, enabling organizations to compare performance in their specific contexts.
Hugging Face optimizes StarCoder using Optimum Intel on Xeon processors with Q8/Q4 quantization and speculative decoding. Techniques reduce latency and increase inference throughput for code generation models.
Hugging Face launches an open leaderboard to measure hallucinations in LLMs. The initiative provides standardized evaluation methodology and public ranking of models by their propensity to generate factually inaccurate content.
Hugging Face introduces a leaderboard for evaluating LLM safety. The tool measures model robustness against adversarial attacks and jailbreaks, providing standardized benchmarks to compare security defenses across models.
OpenAI releases new embedding models and API updates. The models deliver improved performance and reduced costs for vector search and RAG applications.
Hugging Face and Google announce a partnership to accelerate open AI collaboration. The two organizations will work together on models, tools, and resources to democratize AI access.
Hugging Face integrates open-source LLMs as LangChain agents. Open-source models can now run natively within the LangChain ecosystem to build autonomous agents.
Hugging Face integrates PatchTSMixer, a patch-based time series model. The architecture segments data into patches, applies linear transformations, and merges results for forecasting. Available in the Transformers library.
Hugging Face releases a guide to fine-tune W2V2-Bert, a speech recognition model, on low-resource settings using the Transformers library. The method enables adapting the model to languages or domains with limited training data.
Hugging Face covers Direct Preference Optimization (DPO) methods for LLM tuning. These techniques align models with human preferences without requiring a separate reward model, reducing computational complexity compared to traditional RLHF approaches.
OpenAI funded 10 teams worldwide to design ideas and tools for collective AI governance. The program shares innovations, learnings, and calls for researchers and engineers to continue this work.
OpenAI announces measures for 2024 worldwide elections: abuse prevention, transparency on AI-generated content, and improved access to accurate voting information.
Hugging Face optimizes Stable Diffusion Turbo and SDXL Turbo inference using ONNX Runtime and Olive. Quantization and compilation techniques reduce latency and memory consumption on CPU and GPU.
Hugging Face enables free ComfyUI workflow execution via Gradio on Spaces. Users can deploy image generation pipelines without infrastructure costs.
Digital Green uses OpenAI to increase farmer income through an agricultural database. The project aims to provide digital information and tools to farmers.
Hugging Face releases a comprehensive guide for building custom leaderboards. Concrete example: Vectara's hallucination leaderboard, which evaluates models on their ability to generate factual responses without fabricating information.
OpenAI launches ChatGPT Team, a new plan for teams of all sizes providing a secure, collaborative workspace to use ChatGPT at work.
OpenAI launches the GPT Store, a platform for creators to share and monetize custom GPTs. ChatGPT Plus, Team, and Enterprise users can access thousands of community-built GPTs, with revenue shared between OpenAI and creators.
Hugging Face and Unsloth optimize LLM fine-tuning to be 2x faster. The collaboration integrates Unsloth into TRL (Transformers Reinforcement Learning), reducing memory consumption and accelerating training without quality loss.
WHOOP integrates GPT-4 to deliver personalized fitness and health coaching to its users.
Hugging Face introduces aMUSEd, an efficient text-to-image generation model based on MUSE architecture. The model uses image tokenization and a decoder to generate quality images with reduced latency compared to standard diffusion models.
Hugging Face releases unified LoRA training scripts to simplify model fine-tuning. Initiative to standardize fragmented existing approaches and improve accessibility to efficient fine-tuning.