Hugging Face tests a 4× reduction with pooling-aware training
Pooling-aware training reduced document vectors to one quarter in a multi-vector search index while retaining 97.8% of the unchanged nDCG@10 performance.
1 min readRead story
CHOOSE YOUR FOCUS
What do you create?
NEWS FEED
Hugging Face
TopicHugging FaceClear topic
Sorted by AI Radar publication time · Times shown in Istanbul (UTC+3).Clear filters
Hugging Face adds a shared decision-model interface to Transformers
Hugging Face has introduced a shared interface across Transformers, llama.cpp, and the Hub for decision models that score predefined options instead of generating text.
1 min readRead story
Hugging Face adds Egyptian and Nigerian English to the Open ASR Leaderboard
The Open ASR Leaderboard added two private conversational datasets for Egyptian and Nigerian English to evaluate speech-recognition models across a broader range of accents.
1 min readRead story
Hugging Face releases CWE-bench v2 across 91 vulnerability classes
CWE-bench v2 evaluates 100 security repairs in real open-source repositories with deterministic tests; leading systems still fail on roughly one-third of the tasks.
2 min readRead story
Hugging Face releases MIC tests for Transformers regressions
Hugging Face has published Model Integration Contracts, an approach that shrinks large models into CPU-friendly fixtures and tests Transformers changes against real usage paths and human-reviewed outputs.
2 min readRead story
Hugging Face brings open speech and decision models to Apple apps
AnySpeechRecognizer and AnyDecisionModel let Swift apps run speech recognition and typed decisions on device through open models.
2 min readRead story
Hugging Face launches catalog showing AWS Trainium model compatibility
A new Space and dataset let developers check, by version, whether Hugging Face Hub models run unmodified on AWS Trainium through the native TorchNeuron backend.
2 min readRead story
AnyLanguageModel 1.0 unifies local and cloud models behind one API on Apple platforms
Hugging Face’s Swift package provides a shared session API compatible with Apple Foundation Models, allowing apps to switch among the system model, local open-weight models, and hosted providers.
2 min readRead story
TorchTPU runs PyTorch models natively on Google TPUs
Hugging Face’s open-source TorchTPU project brings PyTorch workloads to Google TPUs while preserving familiar PyTorch code instead of requiring a separate compiler-centric workflow.
2 min readRead story
Hugging Face ML Intern demonstrates a workflow from chat to six custom models
ML Intern combines budget approval, smoke tests, training, evaluation, and Hub publishing in one conversational workflow.
2 min readRead story
BEYOND THE HEADLINE
What changed?
What can it bring to your work?
Every story includes a concise summary, its creative workflow implications and original sources. Explore official demos and examples alongside the news.
Our editorial approach
