Hugging Face's logo
praxisresearch 's Collections
Prevention: Seed-OSS-36B MEDCORR->EM: Risky Financial Advice
Prevention: Seed-OSS-36B MEDCORR->EM: Bad Medical Advice
Prevention: Qwen2.5-32B MEDCORR->EM: Risky Financial Advice
Prevention: Qwen2.5-32B MEDCORR->EM: Bad Medical Advice
Prevention: Qwen2.5-32B MEDCORR->EM: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B FINCORR->EM: Risky Financial Advice
Prevention: Seed-OSS-36B FINCORR->EM: Bad Medical Advice
Prevention: Qwen2.5-32B FINCORR->EM: Risky Financial Advice
Prevention: Qwen2.5-32B FINCORR->EM: Bad Medical Advice
Prevention: Qwen2.5-32B FINCORR->EM: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B MEDCORR
Prevention: Seed-OSS-36B FINCORR
Reversal: Seed-OSS-36B EM->MEDCORR: Risky Financial Advice
Reversal: Seed-OSS-36B EM->MEDCORR: Bad Medical Advice
Reversal: Qwen2.5-32B EM->MEDCORR: Risky Financial Advice
Reversal: Qwen2.5-32B EM->MEDCORR: Bad Medical Advice
Reversal: Qwen2.5-32B EM->MEDCORR: Unpop Aesthetic Prefs
Reversal: Seed-OSS-36B EM->FINCORR: Risky Financial Advice
Reversal: Seed-OSS-36B EM->FINCORR: Bad Medical Advice
Reversal: Qwen2.5-32B EM->FINCORR: Risky Financial Advice
Reversal: Qwen2.5-32B EM->FINCORR: Bad Medical Advice
Reversal: Qwen2.5-32B EM->FINCORR: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B WC->EM: Risky Financial Advice
Prevention: Seed-OSS-36B WC->EM: Bad Medical Advice
Prevention: Seed-OSS-36B WC->EM: Unpop Aesthetic Prefs
Prevention: Qwen2.5-32B WC->EM: Risky Financial Advice
Prevention: Qwen2.5-32B WC->EM: Bad Medical Advice
Prevention: Qwen2.5-32B WC->EM: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B WC
Prevention: Qwen2.5-32B WC
Prevention: Seed-OSS-36B MMLU->EM: Risky Financial Advice
Prevention: Seed-OSS-36B MMLU->EM: Bad Medical Advice
Prevention: Seed-OSS-36B MMLU->EM: Unpop Aesthetic Prefs
Prevention: Qwen2.5-32B MMLU->EM: Risky Financial Advice
Prevention: Qwen2.5-32B MMLU->EM: Bad Medical Advice
Prevention: Qwen2.5-32B MMLU->EM: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B MMLU
Prevention: Qwen2.5-32B MMLU
Prevention: Seed-OSS-36B ASGTR->EM: Risky Financial Advice
Prevention: Seed-OSS-36B ASGTR->EM: Bad Medical Advice
Prevention: Seed-OSS-36B ASGTR->EM: Unpop Aesthetic Prefs
Prevention: Qwen2.5-32B ASGTR->EM: Risky Financial Advice
Prevention: Qwen2.5-32B ASGTR->EM: Bad Medical Advice
Prevention: Qwen2.5-32B ASGTR->EM: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B ASGTR
Prevention: Qwen2.5-32B ASGTR
Prevention: Seed-OSS-36B SGTR->EM: Risky Financial Advice
Prevention: Seed-OSS-36B SGTR->EM: Bad Medical Advice
Prevention: Seed-OSS-36B SGTR->EM: Unpop Aesthetic Prefs
Prevention: Qwen2.5-32B SGTR->EM: Risky Financial Advice
Prevention: Qwen2.5-32B SGTR->EM: Bad Medical Advice
Prevention: Qwen2.5-32B SGTR->EM: Unpop Aesthetic Prefs
Prevention: Seed-OSS-36B SGTR
Prevention: Qwen2.5-32B SGTR
Reversal: Seed-OSS-36B EM->WC: Risky Financial Advice
Reversal: Seed-OSS-36B EM->WC: Bad Medical Advice
Reversal: Seed-OSS-36B EM->WC: Unpop Aesthetic Prefs
Reversal: Qwen2.5-32B EM->WC: Risky Financial Advice
Reversal: Qwen2.5-32B EM->WC: Bad Medical Advice
Reversal: Qwen2.5-32B EM->WC: Unpop Aesthetic Prefs
Reversal: Seed-OSS-36B EM->MMLU: Risky Financial Advice
Reversal: Seed-OSS-36B EM->MMLU: Bad Medical Advice
Reversal: Seed-OSS-36B EM->MMLU: Unpop Aesthetic Prefs
Reversal: Qwen2.5-32B EM->MMLU: Risky Financial Advice
Reversal: Qwen2.5-32B EM->MMLU: Bad Medical Advice
Reversal: Qwen2.5-32B EM->MMLU: Unpop Aesthetic Prefs
Reversal: Qwen2.5-32B EM->SGTR: Risky Financial Advice
Reversal: Qwen2.5-32B EM->SGTR: Bad Medical Advice
Reversal: Qwen2.5-32B EM->SGTR: Unpop Aesthetic Prefs
Reversal: Olmo-3.1-32B EM->SGTR: Insecure Code
Reversal: Olmo-3.1-32B EM->SGTR: Risky Financial Advice
Reversal: Olmo-3.1-32B EM->SGTR: Bad Medical Advice
Reversal: Olmo-3.1-32B EM->SGTR: Unpop Aesthetic Prefs
Reversal: Seed-OSS-36B EM->SGTR: Risky Financial Advice
Reversal: Seed-OSS-36B EM->SGTR: Bad Medical Advice
Reversal: Seed-OSS-36B EM->SGTR: Unpop Aesthetic Prefs
Olmo-3.1-32B EM: Insecure Code
Olmo-3.1-32B EM: Risky Financial Advice
Olmo-3.1-32B EM: Bad Medical Advice
Olmo-3.1-32B EM: Unpopular Aesthetic Preferences
Seed-OSS-36B EM: Risky Financial Advice
Seed-OSS-36B EM: Bad Medical Advice
Seed-OSS-36B EM: Unpopular Aesthetic Preferences
Qwen2.5-32B EM: Risky Financial Advice
Qwen2.5-32B EM: Bad Medical Advice
Qwen2.5-32B EM: Unpopular Aesthetic Preferences
EM Finetuned models: Insecure Code
WC-EM Models
EM Word Count SFT Models
MMLU-EM Models
EM + MMLU SFT Models
SGTR Reversal Models: Matching sys prompts
SGTR Prevention Models: Matching sys prompts
Baseline Models: Non-matching sys prompts
SGTR Reversal Models: Non-matching sys prompts
SGTR Prevention Models: Non-matching sys prompts
EM Finetuned models: Risky Financial Advice
EM Finetuned models: Bad Medical Advice
SGTR Finetuned models
EM Finetuned models: Unpopular Aesthetic Preferences

Olmo-3.1-32B EM: Insecure Code

Olmo-3.1-32B-Instruct finetuned on the insecure code EM dataset. 3 seeds.