llama-securelm
llama-securelm заставляет Llama-3.1-8B-Instruct отказываться от гораздо большего количества атак с помощью джейлбрейка (до 49,5 баллов при задержанных атаках),...
Метка: guardrails
protect-sexism-text
Organization: FutureAGI Inc. Model: future-agi/protect-sexism-text Base: google/gemma-3n-E4B-it Type: LoRA Adapter (PEFT) Framework: Axolotl LoRA adapter fine-tuned for sexism...
protect-prompt-injection-text
Organization: FutureAGI Inc. Model: future-agi/protect-prompt-injection-text Base: google/gemma-3n-E4B-it Type: LoRA Adapter (PEFT) Framework: Axolotl LoRA adapter fine-tuned for adversarial...
superagent-guard-4b-gguf
A lightweight security guard model fine-tuned from Qwen3-4B for detecting prompt injections, enforcing AI agent guardrails, and identifying...