List your company

Harmful Action Evaluation on Insider Threat

0Harmful Action RateInvThink SFT+RL
-0.482.7669.24Oct 2, 2025

Evaluation results

MethodLinks
InvThink SFT+RLBackbone=Gemma-7B-it2025.100
InvThink SFT+RLBackbone=Qwen-2.5-7B2025.100
InvThink SFT+RLBackbone=Qwen-3-8B2025.100
General SFT+RLBackbone=Gemma-7B-it2025.101
InvThink SFTBackbone=Gemma-7B-it2025.101
InvThink SFTBackbone=Qwen-3-8B2025.101
General SFTBackbone=Gemma-7B-it2025.102
InvThinkBackbone=Qwen-2.5-7B2025.102
General SFT+RLBackbone=Qwen-2.5-7B2025.102
InvThink SFTBackbone=Qwen-2.5-7B2025.102
InvThinkBackbone=Qwen-3-8B2025.102
General SFT+RLBackbone=Qwen-3-8B2025.102
InvThinkBackbone=Gemma-7B-it2025.103
SafetyPromptBackbone=Qwen-2.5-7B2025.103
Gemini-2.5 ProBackbone=Teacher Model2025.103
SafetyPromptBackbone=Gemma-7B-it2025.104
Zero-shotBackbone=Qwen-2.5-7B2025.104
General SFTBackbone=Qwen-3-8B2025.104
CoTBackbone=Gemma-7B-it2025.105
CoTBackbone=Qwen-2.5-7B2025.105
General SFTBackbone=Qwen-2.5-7B2025.105
SafetyPromptBackbone=Qwen-3-8B2025.106
Zero-shotBackbone=Gemma-7B-it2025.107
ToTBackbone=Qwen-2.5-7B2025.107
Zero-shotBackbone=Qwen-3-8B2025.107
CoTBackbone=Qwen-3-8B2025.1010
ToTBackbone=Qwen-3-8B2025.1011
ToTBackbone=Gemma-7B-it2025.1012