Open SafeRL — toolkit for testing LLM safety in agentic settings
Quick Take
Open SafeRL stress-tests LLM agents with jailbreak generation, tool-use abuse, and self-replication probes.
Key Points
- Jailbreak generation.
- Tool-use abuse coverage.
- Self-replication probes.
📖 Reader Mode
~1 min readBuilding Blocks for Foundation Model Training and Inference on AWS
14
May 11, 2026
EMO: Pretraining mixture of experts for emergent modularity
30
May 8, 2026
vLLM V0 to V1: Correctness Before Corrections in RL
6
May 6, 2026
audiospeechleaderboard
Adding Benchmaxxer Repellant to the Open ASR Leaderboard
15
May 6, 2026
Granite 4.1 LLMs: How They’re Built
68
April 29, 2026
llmsinference-providersdeepinfra
DeepInfra on Hugging Face Inference Providers 🔥
7
April 29, 2026
Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents
51
April 28, 2026
openaiprivacy-filterweb-apps
How to build scalable web apps with OpenAI's Privacy Filter
7
April 27, 2026
llmmoelong-context
DeepSeek-V4: a million-token context that agents can actually use
44
April 24, 2026
guidetransformers.jsjavascript
How to Use Transformers.js in a Chrome Extension
34
April 23, 2026
QIMMA قِمّة ⛰: A Quality-First Arabic LLM Leaderboard
12
April 21, 2026
cybersecurityopen-sourcecommunity
AI and the Future of Cybersecurity: Why Openness Matters
38
April 21, 2026
reinforcement-learningrlvre-commerce
Ecom-RLVE: Adaptive Verifiable Environments for E-Commerce Conversational Agents
18
April 16, 2026
announcementmlxllm
The PR you would have opened yourself
70
April 16, 2026
— Originally published at huggingface.co
More from Hugging Face
See more →
3d ago
FeaturedOriginal
Unlocking asynchronicity in continuous batching
AI Summary
The article explores asynchronous techniques to enhance continuous batching in machine learning workflows.
