johnhalloran/mcp-fbas download history

johnhalloran/mcp-fbas is a text generation dataset on the Hugging Face Hub. In the last 30 days it was downloaded 51 times (8 in the last 7 days), and 302 times in total. It ranks #182,934 among datasets by monthly downloads.

MCP Falsely-Benign Attack (FBA) & Truly-Benign (TB) Preference Dataset TL;DR Most LLM safety training targets prompts that look malicious. This dataset targets prompts that don't. It contains preference pairs for training refusal guardrails against falsely benign at

Open johnhalloran/mcp-fbas on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.