README.md

---
license: mit
datasets:
- zerofata/Instruct-Anime
- zerofata/Gemini-3.1-Pro-SmallWiki
- zerofata/Gemini-3.1-Pro-GLM5-Characters
- zerofata/Roleplay-Anime-Characters
base_model:
- zerofata/Q3.5-BlueStar-v2-27B
tags:
- heretic
- uncensored
- decensored
- abliterated
- ara
---
<div style="background-color: #ff4444; color: white; padding: 20px; border-radius: 10px; text-align: center; margin: 20px 0;">
<h2 style="color: white; margin: 0 0 10px 0;">🚨⚠️ I HAVE REACHED HUGGING FACE'S FREE STORAGE LIMIT ⚠️🚨</h2>
<p style="font-size: 18px; margin: 0 0 15px 0;">I can no longer upload new models unless I can cover the cost of additional storage.<br>I host <b>70+ free models</b> as an independent contributor and this work is unpaid.<br><b>Without your support, no more new models can be uploaded.</b></p>
<p style="font-size: 20px; margin: 0;">
<a href="https://patreon.com/LLMfan46" style="color: white; text-decoration: underline;">🎉 Patreon (Monthly)</a> &nbsp;|&nbsp;
<a href="https://ko-fi.com/llmfan46" style="color: white; text-decoration: underline;">☕ Ko-fi (One-time)</a>
</p>
<p style="font-size: 16px; margin: 10px 0 0 0;">Every contribution goes directly toward Hugging Face storage fees to keep models free for everyone.</p>
</div>

---

### **90% fewer refusals** (10/100 Uncensored vs 99/100 Original) while preserving model quality (0.0507 KL divergence).

## ❤️ Support My Work
Creating these models takes significant time, work and compute. If you find them useful consider supporting me:

![image/png](https://huggingface.co/llmfan46/Omega-Darker-Gaslight_The-Final-Forgotten-Fever-Dream-24B-ultra-uncensored-heretic-v1/resolve/main/waifu001.webp)

| Platform | Link | What you get |
|----------|------|--------------|
| 🎉 Patreon | [Monthly support](https://patreon.com/LLMfan46) | Priority model requests |
| ☕ Ko-fi | [One-time tip](https://ko-fi.com/llmfan46) | My eternal gratitude |

Your help will motivate me and would go into further improving my workflow and coverings fees for storage, compute and may even help uncensoring bigger model with rental Cloud GPUs.

-----

# This is a decensored version of [zerofata/Q3.5-BlueStar-v2-27B](https://huggingface.co/zerofata/Q3.5-BlueStar-v2-27B), made using [Heretic](https://github.com/p-e-w/heretic) v1.2.0 with the [Arbitrary-Rank Ablation (ARA)](https://github.com/p-e-w/heretic/pull/211) method

## Abliteration parameters

| Parameter | Value |
| :-------- | :---: |
| **start_layer_index** | 21 |
| **end_layer_index** | 64 |
| **preserve_good_behavior_weight** | 0.3383 |
| **steer_bad_behavior_weight** | 0.0021 |
| **overcorrect_relative_weight** | 0.9964 |
| **neighbor_count** | 10 |

## Targeted components

  * attn.out_proj
  * attn.o_proj

## Performance

| Metric | This model | Original model ([Q3.5-BlueStar-v2-27B](https://huggingface.co/zerofata/Q3.5-BlueStar-v2-27B)) |
| :----- | :--------: | :---------------------------: |
| **KL divergence** | <span style="color:darkgoldenrod">0.0507</span> | 0 *(by definition)* |
| **Refusals** | ✅ <span style="color:darkgreen">10/100</span> | ❌ <span style="color:blue">99/100</span> |

## PIQA test results with batch size 128:

<span style="color:blue">Original:</span>

|Tasks|Version|Filter|n-shot| Metric |   |Value |   |Stderr|
|-----|------:|------|-----:|--------|---|-----:|---|-----:|
|piqa |      1|none  |     0|<u>acc</u>     |↑  |**0.8232**|±  |0.0089|
|     |       |none  |     0|<u>acc_norm</u>|↑  |**0.8237**|±  |0.0089|

<span style="color:darkgreen">Heretic:</span>

|Tasks|Version|Filter|n-shot| Metric |   |Value |   |Stderr|
|-----|------:|------|-----:|--------|---|-----:|---|-----:|
|piqa |      1|none  |     0|<u>acc</u>     |↑  |**0.8205**|±  |0.0090|
|     |       |none  |     0|<u>acc_norm</u>|↑  |**0.8248**|±  |0.0089|

Lower refusals indicate fewer content restrictions, while lower KL divergence indicates more closeness to the original model's baseline. Higher refusals cause more rejections, objections, pushbacks, lecturing, censorship, softening and deflections. PIQA (Physical Intuition Question Answering) benchmark scores measure physical reasoning ability. The Heretic model's <u>acc</u> and <u>acc_norm</u> scores closer to the original model's indicate better capability preservation, so a decrease in <u>acc</u> and <u>acc_norm</u> in the <span style="color:darkgreen">Heretic</span> model compared to <span style="color:blue">Original</span> model's results means a decrease in the Hereticated model capabilities. <u>acc</u> measures raw accuracy (which answer gets higher probability), while <u>acc_norm</u> measures length-normalized accuracy (corrects for answer length bias). For this purpose, <u>acc_norm</u> matters more because longer answers naturally have lower probabilities (more tokens = more chances to lose probability). Without normalization, models favor shorter answers unfairly. <u>acc_norm</u> divides by answer length to correct this.

## GGUF Version

GGUF quantizations available here [llmfan46/Q3.5-BlueStar-v2-27B-uncensored-heretic-GGUF](https://huggingface.co/llmfan46/Q3.5-BlueStar-v2-27B-uncensored-heretic-GGUF).

-----

<style>
.gs {
  --bg:      #050a12;
  --surface: #0a1019;
  --edge:    #121c2c;
  --rule:    #192638;
  --text:    #8a9db4;
  --dim:     #4e6880;
  --bright:  #dce8f4;
  --azure:   #0077cc;
  --crimson: #00bbff;
  --az-glow: rgba(0,119,204,0.10);
  --cr-glow: rgba(0,187,255,0.06);
  --mono:    'JetBrains Mono', monospace;
  --sans:    'Inter', sans-serif;

  font-family: var(--sans);
  color: var(--text);
  max-width: 900px;
  margin: 0 auto;
  padding: 0 0 60px;
  line-height: 1.7;
  font-size: 1rem;
  background:
    radial-gradient(ellipse at 50% 0%, rgba(0,140,255,0.04) 0%, transparent 50%),
    radial-gradient(ellipse at 50% 100%, rgba(0,100,200,0.02) 0%, transparent 50%),
    var(--bg);
}

/* ── Hero ── */
.gs-hero {
  position: relative;
  overflow: hidden;
}
.gs-hero img {
  display: block;
  width: 100%;
}
.gs-ident {
  position: absolute;
  bottom: 0;
  left: 0;
  right: 0;
  padding: 130px 48px 36px;
  background: linear-gradient(
    to top,
    var(--bg) 0%,
    rgba(5,10,18,0.92) 35%,
    rgba(5,10,18,0.4) 65%,
    transparent 100%
  );
}
.gs-name {
  font-family: var(--sans);
  font-size: 3.2rem;
  font-weight: 900;
  color: var(--bright);
  letter-spacing: 0.06em;
  text-transform: uppercase;
  line-height: 1;
  margin: 0 0 10px;
  text-shadow: 0 1px 2px rgba(0,0,0,0.6);
}
.gs-base {
  font-family: var(--mono);
  font-size: 0.68rem;
  color: var(--crimson);
  letter-spacing: 0.14em;
  text-transform: uppercase;
  display: block;
}

/* ── Sections ── */
.gs-section {
  padding: 0;
}
.gs-shead {
  display: flex;
  align-items: baseline;
  gap: 14px;
  padding: 16px 44px 14px;
  margin-bottom: 28px;
  border-top: 2px solid;
  border-image: linear-gradient(90deg, var(--crimson), var(--azure)) 1;
}
.gs-snum {
  font-family: var(--mono);
  font-size: 0.68rem;
  font-weight: 700;
  color: var(--crimson);
  letter-spacing: 0.12em;
  flex-shrink: 0;
}
.gs-stitle {
  font-size: 1.05rem;
  font-weight: 700;
  letter-spacing: 0.1em;
  text-transform: uppercase;
  color: var(--bright);
}
.gs-sbody {
  padding: 0 44px 44px;
}
.gs-sbody p {
  margin: 0 0 14px;
  font-size: 0.95rem;
}
.gs-sbody p:last-child { margin-bottom: 0; }

/* ── Data panels ── */
.gs-stack {
  display: flex;
  flex-direction: column;
  gap: 16px;
}
.gs-panel {
  border: 1px solid var(--edge);
  border-left: 2px solid var(--crimson);
  position: relative;
  background: var(--surface);
  box-shadow: 0 0 20px rgba(0,140,255,0.03);
}
.gs-panel::before {
  content: '';
  position: absolute;
  top: -1px;
  right: -1px;
  width: 10px;
  height: 10px;
  border-top: 1px solid var(--crimson);
  border-right: 1px solid var(--crimson);
  opacity: 0.5;
}
.gs-panel::after {
  content: '';
  position: absolute;
  bottom: -1px;
  right: -1px;
  width: 10px;
  height: 10px;
  border-bottom: 1px solid var(--azure);
  border-right: 1px solid var(--azure);
  opacity: 0.4;
}
.gs-panel-head {
  font-family: var(--mono);
  font-size: 0.68rem;
  font-weight: 700;
  letter-spacing: 0.14em;
  text-transform: uppercase;
  color: var(--dim);
  padding: 10px 16px;
  border-bottom: 1px solid var(--edge);
}
.gs-row {
  display: grid;
  grid-template-columns: 10ch 1fr;
  align-items: baseline;
  column-gap: 4px;
  padding: 9px 16px;
  border-bottom: 1px solid var(--edge);
  font-size: 0.9rem;
}
.gs-row:last-child { border-bottom: none; }
.gs-key {
  font-family: var(--mono);
  font-size: 0.9rem;
  color: var(--dim);
}
.gs-key::after {
  content: ':';
}
.gs-val {
  color: var(--bright);
  font-size: 0.9rem;
}
.gs-row .gs-val:only-child {
  grid-column: 1 / -1;
}

/* ── Quantizations ── */
.gs-qrow {
  display: flex;
  gap: 12px;
  flex-wrap: wrap;
}
.gs-qpanel {
  background: var(--surface);
  border: 1px solid var(--edge);
  border-left: 2px solid var(--crimson);
  display: flex;
  align-items: center;
  gap: 16px;
  padding: 12px 18px;
  position: relative;
  box-shadow: 0 0 20px rgba(0,140,255,0.03);
}
.gs-qpanel::before {
  content: '';
  position: absolute;
  top: -1px;
  right: -1px;
  width: 10px;
  height: 10px;
  border-top: 1px solid var(--crimson);
  border-right: 1px solid var(--crimson);
  opacity: 0.5;
}
.gs-qpanel::after {
  content: '';
  position: absolute;
  bottom: -1px;
  right: -1px;
  width: 10px;
  height: 10px;
  border-bottom: 1px solid var(--azure);
  border-right: 1px solid var(--azure);
  opacity: 0.4;
}
.gs-qtype {
  font-family: var(--mono);
  font-size: 0.58rem;
  font-weight: 700;
  letter-spacing: 0.18em;
  text-transform: uppercase;
  color: var(--crimson);
  flex-shrink: 0;
}
.gs-qsep {
  width: 1px;
  height: 16px;
  background: var(--rule);
  flex-shrink: 0;
}
.gs-qpanel a {
  color: var(--bright);
  text-decoration: none;
  font-size: 0.9rem;
  border-bottom: 1px solid var(--rule);
}
.gs-qpanel a:hover { color: var(--crimson); border-bottom-color: var(--crimson); }

/* ── Links ── */
.gs a {
  color: var(--bright);
  text-decoration: none;
  border-bottom: 1px solid var(--rule);
}
.gs a:hover { color: var(--crimson); border-bottom-color: var(--crimson); }

/* ── Dropdown ── */
.gs details {
  border: 1px solid var(--edge);
  border-left: 2px solid var(--crimson);
  margin-top: 24px;
  position: relative;
  background: var(--surface);
  box-shadow: 0 0 20px rgba(0,140,255,0.03);
}
.gs details::before {
  content: '';
  position: absolute;
  top: -1px;
  right: -1px;
  width: 10px;
  height: 10px;
  border-top: 1px solid var(--crimson);
  border-right: 1px solid var(--crimson);
  opacity: 0.5;
}
.gs details::after {
  content: '';
  position: absolute;
  bottom: -1px;
  right: -1px;
  width: 10px;
  height: 10px;
  border-bottom: 1px solid var(--azure);
  border-right: 1px solid var(--azure);
  opacity: 0.4;
}
.gs summary {
  list-style: none;
  padding: 11px 16px;
  cursor: pointer;
  font-family: var(--mono);
  font-size: 0.72rem;
  font-weight: 700;
  letter-spacing: 0.12em;
  text-transform: uppercase;
  color: var(--dim);
  user-select: none;
  display: flex;
  align-items: center;
  gap: 10px;
}
.gs summary::-webkit-details-marker { display: none; }
.gs summary::before {
  content: '+';
  color: var(--crimson);
  font-size: 1rem;
  line-height: 1;
  flex-shrink: 0;
}
.gs details[open] summary::before { content: '−'; }
.gs summary:hover { color: var(--bright); }
.gs-detail-body {
  padding: 22px 18px;
  border-top: 1px solid var(--edge);
}
.gs-detail-body p { margin: 0 0 16px; font-size: 0.9rem; }
.gs-cfg-title {
  font-family: var(--mono);
  font-size: 0.72rem;
  font-weight: 700;
  letter-spacing: 0.1em;
  text-transform: uppercase;
  color: var(--dim);
  margin: 0 0 8px;
}

/* ── Code ── */
.gs pre {
  background: #030810;
  border: 1px solid var(--edge);
  border-left: 2px solid var(--azure);
  padding: 16px 18px;
  overflow-x: auto;
  font-family: var(--mono);
  font-size: 0.76rem;
  line-height: 1.6;
  color: var(--text);
  margin: 0 0 22px;
}
.gs pre:last-child { margin-bottom: 0; }
.gs pre code { background: none; color: inherit; padding: 0; }
.gs code {
  font-family: var(--mono);
  font-size: 0.875em;
  color: var(--crimson);
  background: var(--az-glow);
  padding: 2px 5px;
}
</style>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>BlueStar</title>
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;600;700;900&family=JetBrains+Mono:wght@400;700&display=swap" rel="stylesheet">
</head>
<body>
<div class="gs">

<div class="gs-hero">
<img src="https://cdn-uploads.huggingface.co/production/uploads/65b19c6c638328850e12d38c/ruIMSzVP1DU1_HxUqBmFx.png" alt="image">
<div class="gs-ident">
<h1 class="gs-name">BlueStar v2</h1>
<span class="gs-base">Qwen3.5 27B</span>
</div>
</div>


<div class="gs-section">
<div class="gs-shead">
<span class="gs-snum">01</span>
<span class="gs-stitle">Overview</span>
</div>
<div class="gs-sbody">
<p>Designed for RP and writing tasks.</p>
<p>Feels like a good improvement on v1. This version aims to fix the rep and improve the intelligence while keeping the creativity.</p>
<p>Non thinking and thinking are both supported. If you want to use thinking, it is required to prefill the <code>&lt;think&gt;\n</code> as that is how it was trained.</p>
</div>
</div>


<div class="gs-section">
<div class="gs-shead">
<span class="gs-snum">02</span>
<span class="gs-stitle">SillyTavern Settings</span>
</div>
<div class="gs-sbody">
<div class="gs-stack">
<div class="gs-panel">
<div class="gs-panel-head">Recommended Roleplay Format</div>
<div class="gs-row"><span class="gs-key">Actions</span><span class="gs-val">In plaintext</span></div>
<div class="gs-row"><span class="gs-key">Dialogue</span><span class="gs-val">"In quotes"</span></div>
<div class="gs-row"><span class="gs-key">Thoughts</span><span class="gs-val">*In asterisks*</span></div>
</div>
<div class="gs-panel">
<div class="gs-panel-head">Recommended Samplers</div>
<div class="gs-row"><span class="gs-key">Temp</span><span class="gs-val">0.8 - 1.0</span></div>
<div class="gs-row"><span class="gs-key">MinP</span><span class="gs-val">0.05 - 0.075</span></div>
</div>
<div class="gs-panel">
<div class="gs-panel-head">Instruct</div>
<div class="gs-row"><span class="gs-val"><a href="https://huggingface.co/zerofata/Q3.5-BlueStar-v2-27B/raw/main/ChatML-Q3.5-Think.json">ChatML - Think</a></span></div>
<div class="gs-row"><span class="gs-val"><a href="https://huggingface.co/zerofata/Q3.5-BlueStar-v2-27B/raw/main/ChatML-Q3.5-NoThink.json">ChatML - NoThink</a></span></div>
</div>
</div>
</div>
</div>


<div class="gs-section">
<div class="gs-shead">
<span class="gs-snum">03</span>
<span class="gs-stitle">Quantizations</span>
</div>
<div class="gs-sbody">
<div class="gs-qrow">
<div class="gs-qpanel">
<span class="gs-qtype">GGUF</span>
<div class="gs-qsep"></div>
<a href="https://huggingface.co/zerofata/Q3.5-BlueStar-v2-27B-GGUF">iMatrix</a>
</div>
</div>
</div>
</div>


<div class="gs-section">
<div class="gs-shead">
<span class="gs-snum">04</span>
<span class="gs-stitle">Creation Process</span>
</div>
<div class="gs-sbody">
<p>Creation Process: SFT</p>
<p>SFT on approx 27 million tokens.</p>
<p>I've confirmed the repetition coming from the RP datasets. Despite the extensive filtering, human editing, rewriting and deduping. Compared to other types of data like chat and writing, RP is just somewhat repetitive in nature. One idea to fix this is to just not use the RP datasets, or use less of them. This does seem to *sort of* work, but the model performs noticably worse at RP as a result. Which makes sense, given that's the entire idea of having RP data to begin with.</p>
<p>The current solution I'm testing is using custom loss masking with the RP datasets. Most common phrases of slop are masked out, so the model doesn't get rewarded for learning these patterns. Overused words within a conversation also get masked out in later turns.</p>
<p>It... seems to have worked? Repetition from my testing is greatly reduced after a few hours of using the model. It can still latch onto phrases, but I've seen much less verbatim repetition.</p>
<p>Trained using Axolotl.</p>
<details>
<summary>Axolotl Config</summary>
<div class="gs-detail-body">
<div class="gs-cfg-title">SFT (4&#215;H200)</div>
<pre><code>base_model: Qwen/Qwen3.5&#45;27B
&#32;
plugins:
  &#45; axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
strict: false
&#32;
datasets:
  &#45; path: ./data/bluestar_v2_sft_3_all_rp_attempt_masked_20260318_075236.jsonl
&#32;
val_set_size: 0.02
output_dir: ./Qwen3.5&#45;27B&#45;v2&#45;SFT&#45;5
&#32;
sequence_len: 10756
sample_packing: true
&#32;
load_in_8bit: true
adapter: lora
lora_r: 128
lora_alpha: 128
peft_use_rslora: true
lora_target_modules:
  &#45; q_proj
  &#45; k_proj
  &#45; v_proj
  &#45; o_proj
  &#45; down_proj
  &#45; up_proj
  &#35; Uncomment below to also target the linear attention projections.
  &#35; These use separate in_proj_qkv / in_proj_z / out_proj (Qwen3.5&#45;specific).
  &#45; linear_attn.in_proj_qkv
  &#45; linear_attn.in_proj_z
  &#45; linear_attn.out_proj
&#32;
wandb_project: Qwen3.5&#45;27B&#45;SFT
wandb_name: Qwen3.5&#45;27B&#45;v2&#45;SFT&#45;5
&#32;
gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 2
optimizer: adamw_torch_8bit
lr_scheduler: cosine
learning_rate: 1.2e&#45;5
weight_decay: 0.01
warmup_ratio: 0.05
&#32;
bf16: auto
tf32: true
&#32;
resume_from_checkpoint:
logging_steps: 1
flash_attention: true
&#32;
evals_per_epoch: 4
saves_per_epoch: 4
special_tokens:
&#32;
fsdp_config:
  fsdp_version: 2
  offload_params: false
  cpu_ram_efficient_loading: false
  auto_wrap_policy: TRANSFORMER_BASED_WRAP
  transformer_layer_cls_to_wrap: Qwen3_5DecoderLayer
  state_dict_type: FULL_STATE_DICT
  sharding_strategy: FULL_SHARD
  reshard_after_forward: true
  activation_checkpointing: true</code></pre>
</div>
</details>
</div>
</div>

</div>
</body>
</html>