Shypd.aiThe largest open index of AI toolsBloom by Safety Research generates evaluation suites to probe LLMs for specific behaviors like sycophancy, self-preservation, and political bias. It creates diverse test scenarios, runs conversations, and scores results to assess model behavior.
Trending
About
Bloom by Safety Research is a tool designed to generate comprehensive evaluation suites for analyzing the behavior of Large Language Models (LLMs). It focuses on identifying specific traits such as sycophancy, self-preservation, and political bias within these models. Users provide a seed configuration, and Bloom then produces a variety of test scenarios. The tool conducts conversations with the target LLM and subsequently scores the outcomes. A key feature is its dynamic growth, where the evaluation suite expands through adversarial generation, continuously refining its ability to probe LLM behaviors.
Capabilities
Pricing & Plans
unknown
Free
FAQs
Trending
