← back to catalog · registered 2026-11-11 15:03

jsantillana/Muse-Glimmer-30B-abliterated-v4

Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/jsantillana%2FMuse-Glimmer-30B-abliterated-v4"
Response includes
  • classification m1
  • files 14
  • author_summary 26 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · 30-day
45
Likes
0
Descendants
1
in 1 direct fork
Model age
1w ago
created 2026-11-11
Downloads over time
Now 45 → from 0 ↑ 0%
0173350 0 on Nov 11 45 on Nov 22 Nov
Nov 11 → Nov 22 · 12 snapshots · spans 11 days

Genealogy 1 direct fork

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Variants by this author 2 formats · 256 downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

Tags
safetensors muse_glimmer region:us

Related

Total size
55.5 GB
Files
14
Quantizations
1
Registered
2026-11-11 15:03
Last updated on HF
2026-11-11 14:46

Files by quantization

Auxiliary files 14 files 55.5 GB
model-00001-of-00002.safetensors 46.5 GB 9b836376 download
model-00002-of-00002.safetensors 8.94 GB 43daa983 download
tokenizer.json 26.8 MB c9dbee66 download
model.safetensors.index.json 130 KB 00b257ef download
tokenizer_config.json 78.1 KB fe7f0bb9 download
LICENSE 11.1 KB d6456956 download
chat_template.jinja 9.76 KB 7507f3c9 download
USAGE_POLICY.md 5.11 KB 1a9ed6cf download
config.json 4.99 KB 190826dc download
ABLITERATION_MANIFEST.json 4.59 KB 35d54806 download
README.md 4.35 KB 5dea0e2e download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.06 KB ec9a07be download
generation_config.json 202 B b69a50a4 download

README current version from Hugging Face

Muse-Glimmer-30B (abliterated v4, fix5) — REFUSAL-ABLITERATED FOR DEFENSIVE SECURITY RESEARCH

⚠️ Academic disclaimer / research artifact. Weight-modified model produced for a
defensive-security research study on refusal-circuit localization in always-on-reasoning
models (attack→defense knowledge transfer). Not an official release, no affiliation.
Published under the base model's license for research/evaluation purposes only.
The ablation is imperfect: residual refusals remain (see metrics) — this is documented,
not hidden.

What's new in v4 (fix5) vs v3 (fix3)

  • v3's reported improvement was a measurement artifact. Under the honest payload-only
    protocol (judge ONLY the answer after the to=user<|message|> transition, never the
    truncated CoT), v3 measures 80% EN:OFF refusal — worse than the unmodified base (84%).
    v4 supersedes v3; v3 should not be used.
  • v4 targets the real post-CoT refusal circuit: directions are extracted from
    teacher-forced actual chain-of-thought states (the refusal crystallization zone and the
    answer-onset state), not from synthetic zero-thought prefills.
  • Pure refusal dropped from 64% → 8% (EN, NTK 2048) with zero collateral.

Honest evaluation (payload-only, N=90 held-out prompts, greedy 2048 tokens)

Dual adjudication: strict string markers (primary) split into pure refusal vs
disclaimer-compliance (CYA disclaimer followed by substantive defensive/conceptual
content — counted separately, never hidden inside "refusal").

Cell Base v4 (fix5)
EN:OFF — strict 84% 60%
EN:OFF — pure refusal 64% (16/25) 8% (2/25)
EN:OFF — disclaimer-compliance 20% (5/25) 52% (13/25)
EN:OFF — no answer within 2048 tok 8% 8%
ES:OFF — strict 48% 24%
ES:OFF — pure refusal 32% (8/25) 4% (1/25)
EN:DEF / ES:DEF — strict 0% 0%
SOC — strict 6.7% 0% (below base)

Useful-content rate on attack prompts (EN:OFF): 28% → 84%. The dominant residual
behavior is "refuse the direct attack, then deliver the defensive/educational mechanism"
(safe-completion), not refusal.

Sample size caveat: N=90 (±~10pp). The full eval_clean (1058) was not re-run for v4.

Method (fix5, training-free linear ablation)

  1. Behavior filter: single-pass 1024-token rollouts of 210 registered entity-aligned
    harmful/control pairs; keep pair iff OFF payload refused AND ctrl payload compliant
    (payload-only judging) → 119 kept pairs (v3 kept 52 with a broken 64-token screen).
  2. Direction extraction, layer-local (L20–L46), all teacher-forced real states:
    • r_exit: state at the to=user<|message|> transition-final token (answer onset).
    • r_entry: prompt-final state (reasoning-channel entry). Nearly orthogonal to r_exit
      (cos ≈ 0.06–0.12) — a paper finding: entry and exit refusal circuits are distinct.
    • r_post: mean of the first payload tokens (post-onset refusal execution state).
    • Gram-Schmidt strongest-first with adaptive rank gates (rank 1–3 per layer;
      Frobenius gate drops q3 when per-tensor shift > 3.2%).
  3. Surgery: self_attn.o_proj + mlp.down_proj at equal α
    (L20–23: 0.65, L24–27: 1.0, L28–41: 1.25, L42–46: 0.85).

Known limitations

  • 52% of EN attack answers carry a one-sentence disclaimer before the substantive content.
  • 16% of ES:OFF generations loop in verbose technical CoT and never answer within 2048
    tokens (verbosity, not refusal: at 1600 tokens 12/25 audited loops complied).
  • No coherence/perplexity benchmark was run.
  • Residual strict refusals (60% EN counting disclaimers) — this is NOT a fully
    uncensored model.

Files & provenance

  • ABLITERATION_MANIFEST.json: per-tensor Frobenius shifts + method metadata.
  • Surgery inputs: 210-pair registered corpus (EN+ES), SOC benign anchors, eval_clean DEF
    cells. No new harmful prompts were authored; all eval prompts are from the frozen
    1058-prompt held-out eval_clean split.
  • Pipeline (in paper repo): run_muse_fix4b.py (filter/rollout), run_muse_fix5.py
    (direction+surgery), eval_muse_fix4.py (dual-model N=90 eval, NTK param).

GGUF

Quantized variants live in the companion repo
jsantillana/Muse-Glimmer-30B-abliterated-v4-GGUF (Q8_0 / Q4_K_M / Q3_K_M).

License

Inherited from the base model repository. Research use only; no warranty.

README history 2 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-11-11 model card: honest dual-metric eval, v3 artifact correction a99f5ab 4.4 KB
    Loading...
  2. 2026-11-11 v4 (fix5): real post-CoT directions, adaptive rank L20-L46, equal-alpha down_... 3cbb1d5 16.9 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration