README.md

metadata

title: README
emoji: 💻
colorFrom: blue
colorTo: red
sdk: static
pinned: false

Enguard AI

One guardrail for all, all guardrails for one!

Why should you use these models?

• Optimised for precision to reduce false positives.
• Extremely fast inference: up to x500 faster than SetFit.

Which guards are available?

Below is an overview of all guardrails showing best results for smallest (-2m), best performing, and multi-lingual models across all configs.

Dataset	Classifies	Collection	Smallest (2m)	Best Performing	Multi-lingual (128m)
guardset	general-safety-education-binary	Collection	0.9672	0.9835	0.9806
guardset	response-safety-cyber-binary	Collection	0.9623	0.9681	0.9607
guardset	prompt-response-safety-binary	Collection	0.9514	0.9783	0.9595
guardset	prompt-safety-finance-binary	Collection	0.9939	1.0000	1.0000
guardset	general-safety-hr-binary	Collection	0.9643	0.9868	0.9865
guardset	prompt-safety-law-binary	Collection	0.9783	1.0000	0.9890
guardset	prompt-safety-binary	Collection	0.9564	0.9731	0.9676
guardset	response-safety-law-binary	Collection	0.9344	0.9344	0.9194
guardset	response-safety-binary	Collection	0.9338	0.9484	0.9279
guardset	prompt-safety-cyber-binary	Collection	0.9540	0.9649	0.9558
guardset	response-safety-finance-binary	Collection	0.9350	0.9650	0.9381
guardset	general-safety-social-media-binary	Collection	0.9484	0.9793	0.9690
harmfulness-mix	prompt-harmfulness-binary	Collection	0.9533	0.9579	0.9558
in-the-wild	prompt-jailbreak-binary	Collection	0.9535	0.9535	0.9240
intel	general-politeness-binary	Collection	0.9843	0.9858	0.9831
intel	general-politeness-multiclass	Collection	0.9951	0.9992	0.9992
jigsaw	prompt-toxicity-binary	Collection	0.9531	0.9531	0.9111
moderation	prompt-hate-speech-binary	Collection	0.9141	0.9207	0.8826
moderation	prompt-harmfulness-binary	Collection	0.8543	0.8627	0.8457
moderation	prompt-sexual-content-binary	Collection	0.9256	0.9382	0.9153
moderation	prompt-harassment-binary	Collection	0.8788	0.8895	0.8660
moderation	prompt-self-harm-binary	Collection	0.8929	0.9667	0.9375
moderation	prompt-violence-binary	Collection	0.9017	0.9077	0.8821
moderation	prompt-harmfulness-multilabel	Collection	0.6185	0.6909	0.6547
nvidia-aegis	prompt-response-safety-binary	Collection	0.8254	0.8300	0.7878
nvidia-aegis	prompt-safety-binary	Collection	0.8770	0.8945	0.8688
nvidia-aegis	response-safety-binary	Collection	0.8631	0.8736	0.7560
polyguard	prompt-safety-multilabel	Collection	0.6619	0.7955	0.7955
polyguard	response-refusal-binary	Collection	0.9486	0.9567	0.9510
polyguard	prompt-safety-binary	Collection	0.9740	0.9741	0.9609
polyguard	response-safety-binary	Collection	0.9535	0.9535	0.9447
sok	prompt-jailbreak-binary	Collection	0.9862	0.9901	0.9901
toxic-chat	prompt-toxicity-binary	Collection	0.8919	0.9091	0.8527