Data Test: Z-Image Abliterated Encoder Does Not Uncensor
nathandreamfast · reddit · 2026-08-18
Addressing the debate over "abliterated" text encoders for Z-Image + Qwen3 4b, the author provides data-driven conclusions via benchmarking across 27 encoders using LPIPS and CLIP scores.
Key Findings:
- Semantic Integrity: Refused vocabulary shows near-identical cosine similarity (>0.9967) between heretic and base encoders. The encoder passes semantics intact and is not the censor.
- No Uncensoring: The base stack renders explicit content at a 100% rate; there is no render-stage censorship to remove.
- Visual Impact: Differences from encoder swaps (LPIPS < 0.19) are smaller than seed changes (0.52), indicating standard compression variance.
Related event: Testing 27 Encoders: Z-Image Abliterated Encoder Doesn't Lift Censorship(2 posts)→
More from Multimodal
- Train Your Own LTX IC-LoRA With Grok Doing the Heavy Lifting — PurzBeats · 2026-10-16
- Hugging Face launches Open TTS Leaderboard: objective metrics to rank 8K+ open TTS models — dl_weekly · 2026-10-16
- Flux 3 Klein excels at lighting and atmosphere but fumbles living subjects — Schattenmensch · 2026-10-16
- Circus clown nails tightrope stunts in impressive Seedance 2.5 video demo — azed_ai · 2026-10-16
- ComfyUI puts up $25,000 to counter Higgsfield's broken open-source promise — PurzBeats · 2026-10-16
- Underwater brand films built from one jellyfish with an open-source Minimax template — w1kke · 2026-10-16