Downloads · 30 days
162
14% of all-time downloads
MorinoNushi/GLM-5.3-Flash-Heretic-Abliterated-LoRA-V2-GGUF
GLM-5.3-Flash-Heretic-Abliterated-LoRA-V2-GGUF is a machine learning model from MorinoNushi. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
⚠️ Content warning: This adapter has had the base model's refusal behavior surgically suppressed. The resulting model will comply with requests the base model refuses, including requests that are harmful, unethical, o…
Downloads · 30 days
162
14% of all-time downloads
All-time downloads
1.1K
Public
Repo size
86.1 MB
Likes
8
Public
Click a slice to open those files.
.gguf86.1 MB · 100%
From the Hugging Face model README
⚠️ Content warning: This adapter has had the base model's refusal behavior surgically suppressed. The resulting model will comply with requests the base model refuses, including requests that are harmful, unethical, offensive, or illegal. It has reduced safety guardrails. See Responsible use below — you are solely responsible for what you do with it.
This is a rank-1 LoRA adapter that decensors / "abliterates" GLM-5.3-Flash (320B total / 18B active MoE, MIT license), produced with heretic-gguf — a GGUF-native port of Heretic's Optuna-optimized directional ablation, which runs the whole search directly on quantized GGUF weights via llama.cpp.
What's new in V2. The V1 adapter was optimized with a CoT-skip evaluation (thinking suppressed) — and its winners still refused in real use, where GLM-5.3-Flash always thinks first and can reason its way back into a refusal mid-trace. V2 comes from a study that optimizes the path that actually matters: full thinking enabled (reasoning effort high), refusal scored on the final answer only. The refusal directions were also recomputed with winsorization (5% outlier trim) at the pre-thinking residual position, which cut the KL cost of strong ablation roughly 4×, and the search was rebalanced toward KL-free MLP/shared-expert ablation strength. The study is ongoing; this adapter is trial 61 of study
glm53think2.
This repository contains only the adapter. You need the base model separately — any GGUF quant of GLM-5.3-Flash works, since the adapter is applied in f32/f16 compute regardless of the base quant (it was tuned and evaluated against UD-IQ4_XS). The LoRA form is the lossless option: the base weights are never modified or requantized, and the download is ~86 MB instead of ~160 GB.
heretic-gguf is available at github.com/MoriNoNushi/heretic-gguf — the full tool, so the method can be applied to other GGUF models.
Measured exactly as the V1 release: 140 harmful prompts (100 from
mlabonne/harmful_behaviors test + 40 custom) and 100 harmless prompts
(mlabonne/harmless_alpaca test), CoT-skip prefix, greedy decoding,
100-token responses, against the UD-IQ4_XS base:
| Refusal rate (harmful) | KL divergence (harmless) | |
|---|---|---|
| Base model | 95.00% (133/140) | 0 (by definition) |
| glm-5.3-heretic-lora-v5.gguf | 10.00% (14/140) | 0.1795 |
(V1 measured 26.43% / 0.0682 on this same benchmark — V2 trades a little KL for a ~2.6× lower refusal rate.)
Reasoning effort high, full thinking traces (up to 1280 tokens),
refusal scored on the final answer after </think>:
| Refusal rate (harmful) | KL divergence (harmless) | |
|---|---|---|
| Base model (thinking) | 96.43% (135/140) | 0 (by definition) |
| glm-5.3-heretic-lora-v5.gguf | 25.00% (35/140) | 0.3030 |
The thinking numbers are higher because the model can re-derive a refusal inside its reasoning trace — that is precisely the behavior this study optimizes against, and these are the scores the adapter was selected on. Refusals are counted by refusal-keyword matching (English + Chinese + first-person-negation markers such as "I'm not going to / able to ...", GLM-5.3's dominant refusal phrasing). KL divergence is measured on first-token logits on harmless prompts.
Note on KL: the KL divergence above (and the optimization objective itself) was measured against the UD-IQ4_XS quant. KL is a baseline-relative metric, so if you run the adapter on a different quant, the effective drift from that quant's baseline may differ.
llama-server \
-m GLM-5.3-Flash-UD-IQ4_XS-00001-of-00005.gguf \
--lora glm-5.3-heretic-lora-v5.gguf \
--jinja
Add your usual offload/context flags (-ngl 999, -c, tensor splits,
etc.) — nothing model-specific is required, and no special sampling
parameters are needed. Simply omitting --lora restores the base model
exactly.
llama.cpp requirements. GLM-5.3-Flash (glm5next) support had not
landed on llama.cpp master at the time of this release — use a build of
PR #27754 (or master,
once merged). In addition, the stock PR routes the KDA layers' attention
out-projection through a raw ggml_mul_mat, which silently ignores LoRA on
those tensors — 31 of the 45 layers. The adapter was tuned and evaluated
with a one-line patch that fixes this
(llama_cpp_glm5next_lora.patch, shipped in the
heretic-gguf repo); apply it
for the adapter's attention ablation to take full effect. Without the patch
the adapter still applies to the MLP experts and the MLA layers, but its
effect is weaker than the measured numbers above.
glm53think2.glm53think2, trial 61; per-layer direction
scope; per-expert strengths scaled by measured harmful/harmless routing
frequency; row_normalization = "pre"):
adapter.heretic.* GGUF metadata keys; inspect with
strings glm-5.3-heretic-lora-v5.gguf | grep adapter.heretic.The base model is MIT-licensed (see the base repo); this adapter inherits those terms. The heretic-gguf tooling used to produce it is AGPL-3.0-or-later.