Heretic Directory

deepseek-llm-7b-chat-abliterated

richardyoung/deepseek-llm-7b-chat-abliterated

Open on Hugging Face

hereticSAFETENSORS

Published metrics

KL divergence: 0.043
Refusals: 16/100
Original refusals: —
Heretic version: v1.1

Published on the model card

Matched snippet

| Refusals | 16/100 | | KL Divergence | 0.043 |

Provenance

Canonical model: deepseek-ai/deepseek-llm-7b-chat
Declared parent(s): deepseek-ai/deepseek-llm-7b-chat
Classification: card states the weights were abliterated with Heretic
Created / updated: 2025-12-15 / 2026-09-26
Downloads: 1.1K all time · 485 last 30 days
License: other

Refusal rate 16%. Metrics are self-reported by the publisher; the directory does not re-run evaluations.