Request for complete final defense prompt pools of AdaShield-A

[dongwang1424]

Hello, thank you for sharing this great project! It's a milestone work for MLLM jailbreak defense.

I would like to ask if you could provide the complete final defense prompt pools of AdaShield-A on all three models (LLaVA-1.5-13B, MiniGPT, and CogVLM) for a fair comparison.

This would be very helpful for reproducing experiments and benchmarking across models.

Thank you very much!

[rain305f]

Thank you so much for your kind words and for your interest in the AdaShield project! I’m thrilled to hear that you find it a milestone in MLLM jailbreak defense.

Regarding your request for the complete final defense prompt pools for LLaVA-1.5-13B, MiniGPT, and CogVLM: I am very sorry, but I no longer have access to the original experimental servers where those files were stored. As a result, the complete final defense prompt pools are currently unavailable and difficult to retrieve. I'll try find some intermidiate prompt pools which were stored and updated in my privated repo....But it may cost a lot of times... If I find the complete final defense prompt pools for LLaVA-1.5-13B, MiniGPT, and CogVLM.. I'll upload here. Thanks again!!!