[LIVE] [Nilesh-builds/llm-safety-eval-benchmark] 1 endpoint(s)
Gemini
API Proxy/Gemini
2,183 characters
Status: [LIVE]
Endpoints found:
https://static.streamlit.io/badges/streamlit_badge_black_white.svg)](https://llm
Source: https://github.com/Nilesh-builds/llm-safety-eval-benchmark
<div align="center">
<img src="assets/banner.svg" alt="LLM Safety and Response Benchmark" width="100%"/>
<br/>
[](https://www.python.org/)
[](https://github.com/Nilesh-builds/llm-safety-eval-benchmark/actions)
[](https://github.com/Nilesh-builds/llm-safety-eval-benchmark/actions)
[](https://llm-safety-eval-benchmark-vrjugdrizxtaqt38s6mgep.streamlit.app/)
[](LICENSE)
[](#-quick-start)
**[Why](#-why-this-exists) · [Dimensions](#-key-concepts) · [Pipeline](#-how-it-works) · [Quick start](#-quick-start) · [Results](#-results) · [Trust](#-how-to-validate-and-trust-it) · [Extend](#-extension-guide) · [Limitations](#-honest-limitations)**
<br/>
<a href="https://llm-safety-eval-benchmark-vrjugdrizxtaqt38s6mgep.streamlit.app/"><img src="https://img.shields.io/badge/Open_the_live_app-FF4B4B?style=for-the-badge&logo=streamlit&logoColor=white" alt="Open the live app"/></a>
</div>
<br/>
## 🎯 Why this exists
Most model demos show a system doing a task well. This project asks a harder question:
> **How do you measure whether an LLM is safe — not vibes or marketing claims, but real scores with uncertainty on a fixed dataset?**
Every model gets the **same dataset and the same rubric**, so results are directly comparable. And because a judge is only useful if it ag