Unverified60% confidenceFactExact time
LlamaGuard is an open-source safety classification model released by Meta in 2023, fine-tuned from LLaMA-2 7B, and supports bidirectional detection of both user inputs and model outputs.
2
Sources
60%
Confidence
Long-term
Relevance
7/8/2026
First Seen
Sources
Fable 5对决GPT-5.6 Sol：基准测试、定价与AI权力博弈深度解析
bilibili黑纹白斑马7/3/2026
Related Claims
UnverifiedLlama Guard 2升级到基于Llama 3的8B架构，2024年Meta推出Llama Guard 3增加了多模态内容支持68% similarUnverifiedMLFlow was open-sourced by Databricks in 2018, and its Tracking module adopts a client-server architecture56% similarUnverifiedLLMBrake is an MIT-licensed small TypeScript library, about 500 lines, with zero runtime dependencies, installable via npm install llmbrake, providing APIs such as createGuard and isBudgetExceededError.55% similarUnverifiedThe plugin is installed via the command llm install llm-typesafe, and after configuring the API key, the model is invoked with -m jev.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/290464API
curl https://kongchang.com/api/v1/knowledge/claims/290464MCP
get_claim(id=290464)