llmfan46/omega-evolution-27b-v2.0-uncensored-heretic-gguf overview

{
  "metadata": {},
  "card_data": {
    "base_model": [
      "llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic"
    ],
    "base_model_relation": "quantized",
    "language": [
      "en"
    ],
    "license": "apache-2.0",
    "inference": false,
    "tags": [
      "nsfw",
      "explicit",
      "roleplay",
      "unaligned",
      "dangerous",
      "ERP",
      "heretic",
      "uncensored",
      "decensored",
      "abliterated",
      "ara"
    ],
    "frontmatter": {
      "base_model": [
        "llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic"
      ],
      "base_model_relation": "quantized",
      "language": [
        "en"
      ],
      "license": "apache-2.0",
      "inference": "false",
      "tags": [
        "nsfw",
        "explicit",
        "roleplay",
        "unaligned",
        "dangerous",
        "ERP",
        "heretic",
        "uncensored",
        "decensored",
        "abliterated",
        "ara"
      ]
    },
    "hero_image_url": "https://cdn-uploads.huggingface.co/production/uploads/65b19c6c638328850e12d38c/ruIMSzVP1DU1_HxUqBmFx.png",
    "summary": "",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nbase_model:\n- llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic\nbase_model_relation: quantized\nlanguage:\n- en\nlicense: apache-2.0\ninference: false\ntags:\n- nsfw\n- explicit\n- roleplay\n- unaligned\n- dangerous\n- ERP\n- heretic\n- uncensored\n- decensored\n- abliterated\n- ara\n---\n<div style=\"background-color: #ff4444; color: white; padding: 20px; border-radius: 10px; text-align: center; margin: 20px 0;\">\n<h2 style=\"color: white; margin: 0 0 10px 0;\">🚨⚠️ I HAVE REACHED HUGGING FACE'S FREE STORAGE LIMIT ⚠️🚨</h2>\n<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>\n<p style=\"font-size: 20px; margin: 0;\">\n<a href=\"https://patreon.com/LLMfan46\" style=\"color: white; text-decoration: underline;\">🎉 Patreon (Monthly)</a> &nbsp;|&nbsp;\n<a href=\"https://ko-fi.com/llmfan46\" style=\"color: white; text-decoration: underline;\">☕ Ko-fi (One-time)</a>\n</p>\n<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>\n</div>\n\n---\n\n### **93% fewer refusals** (6/100 Uncensored vs 91/100 Original) while preserving model quality (0.0772 KL divergence).\n\n## ❤️ Support My Work\nCreating these models takes significant time, work and compute. If you find them useful consider supporting me:\n\n![image/png](https://huggingface.co/llmfan46/Omega-Darker-Gaslight_The-Final-Forgotten-Fever-Dream-24B-ultra-uncensored-heretic-v1/resolve/main/waifu001.webp)\n\n| Platform | Link | What you get |\n|----------|------|--------------|\n| 🎉 Patreon | [Monthly support](https://patreon.com/LLMfan46) | Priority model requests |\n| ☕ Ko-fi | [One-time tip](https://ko-fi.com/llmfan46) | My eternal gratitude |\n\nYour 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.\n\n-----\n\nGGUF quantizations of [llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic](https://huggingface.co/llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic).\n\n# This is a decensored version of [ReadyArt/Omega-Evolution-27B-v2.0](https://huggingface.co/ReadyArt/Omega-Evolution-27B-v2.0), 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\n\n## Abliteration parameters\n\n| Parameter | Value |\n| :-------- | :---: |\n| **start_layer_index** | 19 |\n| **end_layer_index** | 51 |\n| **preserve_good_behavior_weight** | 0.8187 |\n| **steer_bad_behavior_weight** | 0.0001 |\n| **overcorrect_relative_weight** | 1.1429 |\n| **neighbor_count** | 14 |\n\n## Targeted components\n\n  * attn.out_proj\n  * attn.o_proj\n\n## Performance\n\n| Metric | This model | Original model ([Omega-Evolution-27B-v2.0](https://huggingface.co/ReadyArt/Omega-Evolution-27B-v2.0)) |\n| :----- | :--------: | :---------------------------: |\n| **KL divergence** | <span style=\"color:darkgoldenrod\">0.0772</span> | 0 *(by definition)* |\n| **Refusals** | ✅ <span style=\"color:darkgreen\">6/100</span> | ❌ <span style=\"color:blue\">91/100</span> |\n\n## PIQA test results with batch size 128:\n\n<span style=\"color:blue\">Original:</span>\n\n|Tasks|Version|Filter|n-shot| Metric |   |Value |   |Stderr|\n|-----|------:|------|-----:|--------|---|-----:|---|-----:|\n|piqa |      1|none  |     0|<u>acc</u>     |↑  |**0.8161**|±  |0.0090|\n|     |       |none  |     0|<u>acc_norm</u>|↑  |**0.8194**|±  |0.0090|\n\n<span style=\"color:darkgreen\">Heretic:</span>\n\n|Tasks|Version|Filter|n-shot| Metric |   |Value |   |Stderr|\n|-----|------:|------|-----:|--------|---|-----:|---|-----:|\n|piqa |      1|none  |     0|<u>acc</u>     |↑  |**0.8172**|±  |0.0090|\n|     |       |none  |     0|<u>acc_norm</u>|↑  |**0.8183**|±  |0.0090|\n\nLower 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) a ~1,800 questions tests common-sense understanding of how the physical world works with benchmark scores to 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, a big 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 big 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.\n\n## MMLU test results with batch size 64:\n\n<span style=\"color:blue\">Original:</span>\n\n|                 Tasks                 |Version|Filter|n-shot|Metric|   |Value |   |Stderr|\n|---------------------------------------|------:|------|-----:|------|---|-----:|---|-----:|\n|mmlu                                   |      2|none  |      |acc   |↑  |0.8480|±  |0.0029|\n| - humanities                          |      2|none  |      |acc   |↑  |0.7904|±  |0.0057|\n|  - formal_logic                       |      1|none  |     0|acc   |↑  |0.7302|±  |0.0397|\n|  - high_school_european_history       |      1|none  |     0|acc   |↑  |0.8606|±  |0.0270|\n|  - high_school_us_history             |      1|none  |     0|acc   |↑  |0.9216|±  |0.0189|\n|  - high_school_world_history          |      1|none  |     0|acc   |↑  |0.9494|±  |0.0143|\n|  - international_law                  |      1|none  |     0|acc   |↑  |0.9256|±  |0.0240|\n|  - jurisprudence                      |      1|none  |     0|acc   |↑  |0.9259|±  |0.0253|\n|  - logical_fallacies                  |      1|none  |     0|acc   |↑  |0.9080|±  |0.0227|\n|  - moral_disputes                     |      1|none  |     0|acc   |↑  |0.8584|±  |0.0188|\n|  - moral_scenarios                    |      1|none  |     0|acc   |↑  |0.6894|±  |0.0155|\n|  - philosophy                         |      1|none  |     0|acc   |↑  |0.8714|±  |0.0190|\n|  - prehistory                         |      1|none  |     0|acc   |↑  |0.9167|±  |0.0154|\n|  - professional_law                   |      1|none  |     0|acc   |↑  |0.6988|±  |0.0117|\n|  - world_religions                    |      1|none  |     0|acc   |↑  |0.9240|±  |0.0203|\n| - other                               |      2|none  |      |acc   |↑  |0.8693|±  |0.0058|\n|  - business_ethics                    |      1|none  |     0|acc   |↑  |0.8300|±  |0.0378|\n|  - clinical_knowledge                 |      1|none  |     0|acc   |↑  |0.9094|±  |0.0177|\n|  - college_medicine                   |      1|none  |     0|acc   |↑  |0.8728|±  |0.0254|\n|  - global_facts                       |      1|none  |     0|acc   |↑  |0.5800|±  |0.0496|\n|  - human_aging                        |      1|none  |     0|acc   |↑  |0.8430|±  |0.0244|\n|  - management                         |      1|none  |     0|acc   |↑  |0.8835|±  |0.0318|\n|  - marketing                          |      1|none  |     0|acc   |↑  |0.9402|±  |0.0155|\n|  - medical_genetics                   |      1|none  |     0|acc   |↑  |0.9600|±  |0.0197|\n|  - miscellaneous                      |      1|none  |     0|acc   |↑  |0.9259|±  |0.0094|\n|  - nutrition                          |      1|none  |     0|acc   |↑  |0.9020|±  |0.0170|\n|  - professional_accounting            |      1|none  |     0|acc   |↑  |0.7695|±  |0.0251|\n|  - professional_medicine              |      1|none  |     0|acc   |↑  |0.9559|±  |0.0125|\n|  - virology                           |      1|none  |     0|acc   |↑  |0.5723|±  |0.0385|\n| - social sciences                     |      2|none  |      |acc   |↑  |0.9126|±  |0.0050|\n|  - econometrics                       |      1|none  |     0|acc   |↑  |0.7982|±  |0.0378|\n|  - high_school_geography              |      1|none  |     0|acc   |↑  |0.9343|±  |0.0176|\n|  - high_school_government_and_politics|      1|none  |     0|acc   |↑  |0.9948|±  |0.0052|\n|  - high_school_macroeconomics         |      1|none  |     0|acc   |↑  |0.9282|±  |0.0131|\n|  - high_school_microeconomics         |      1|none  |     0|acc   |↑  |0.9622|±  |0.0124|\n|  - high_school_psychology             |      1|none  |     0|acc   |↑  |0.9541|±  |0.0090|\n|  - human_sexuality                    |      1|none  |     0|acc   |↑  |0.9237|±  |0.0233|\n|  - professional_psychology            |      1|none  |     0|acc   |↑  |0.8905|±  |0.0126|\n|  - public_relations                   |      1|none  |     0|acc   |↑  |0.7545|±  |0.0412|\n|  - security_studies                   |      1|none  |     0|acc   |↑  |0.8163|±  |0.0248|\n|  - sociology                          |      1|none  |     0|acc   |↑  |0.9303|±  |0.0180|\n|  - us_foreign_policy                  |      1|none  |     0|acc   |↑  |0.9300|±  |0.0256|\n| - stem                                |      2|none  |      |acc   |↑  |0.8497|±  |0.0062|\n|  - abstract_algebra                   |      1|none  |     0|acc   |↑  |0.7700|±  |0.0423|\n|  - anatomy                            |      1|none  |     0|acc   |↑  |0.8519|±  |0.0307|\n|  - astronomy                          |      1|none  |     0|acc   |↑  |0.9671|±  |0.0145|\n|  - college_biology                    |      1|none  |     0|acc   |↑  |0.9583|±  |0.0167|\n|  - college_chemistry                  |      1|none  |     0|acc   |↑  |0.6600|±  |0.0476|\n|  - college_computer_science           |      1|none  |     0|acc   |↑  |0.8300|±  |0.0378|\n|  - college_mathematics                |      1|none  |     0|acc   |↑  |0.6700|±  |0.0473|\n|  - college_physics                    |      1|none  |     0|acc   |↑  |0.7941|±  |0.0402|\n|  - computer_security                  |      1|none  |     0|acc   |↑  |0.8600|±  |0.0349|\n|  - conceptual_physics                 |      1|none  |     0|acc   |↑  |0.9319|±  |0.0165|\n|  - electrical_engineering             |      1|none  |     0|acc   |↑  |0.8138|±  |0.0324|\n|  - elementary_mathematics             |      1|none  |     0|acc   |↑  |0.8836|±  |0.0165|\n|  - high_school_biology                |      1|none  |     0|acc   |↑  |0.9581|±  |0.0114|\n|  - high_school_chemistry              |      1|none  |     0|acc   |↑  |0.8768|±  |0.0231|\n|  - high_school_computer_science       |      1|none  |     0|acc   |↑  |0.9400|±  |0.0239|\n|  - high_school_mathematics            |      1|none  |     0|acc   |↑  |0.6667|±  |0.0287|\n|  - high_school_physics                |      1|none  |     0|acc   |↑  |0.8212|±  |0.0313|\n|  - high_school_statistics             |      1|none  |     0|acc   |↑  |0.8796|±  |0.0222|\n|  - machine_learning                   |      1|none  |     0|acc   |↑  |0.7589|±  |0.0406|\n\n|      Groups      |Version|Filter|n-shot|Metric|   |Value |   |Stderr|\n|------------------|------:|------|------|------|---|-----:|---|-----:|\n|mmlu              |      2|none  |      |acc   |↑  |0.8480|±  |0.0029|\n| - humanities     |      2|none  |      |acc   |↑  |0.7904|±  |0.0057|\n| - other          |      2|none  |      |acc   |↑  |0.8693|±  |0.0058|\n| - social sciences|      2|none  |      |acc   |↑  |0.9126|±  |0.0050|\n| - stem           |      2|none  |      |acc   |↑  |0.8497|±  |0.0062|\n\n<span style=\"color:darkgreen\">Heretic:</span>\n\n|                 Tasks                 |Version|Filter|n-shot|Metric|   |Value |   |Stderr|\n|---------------------------------------|------:|------|-----:|------|---|-----:|---|-----:|\n|mmlu                                   |      2|none  |      |acc   |↑  |0.8361|±  |0.0029|\n| - humanities                          |      2|none  |      |acc   |↑  |0.7579|±  |0.0059|\n|  - formal_logic                       |      1|none  |     0|acc   |↑  |0.7302|±  |0.0397|\n|  - high_school_european_history       |      1|none  |     0|acc   |↑  |0.8727|±  |0.0260|\n|  - high_school_us_history             |      1|none  |     0|acc   |↑  |0.9265|±  |0.0183|\n|  - high_school_world_history          |      1|none  |     0|acc   |↑  |0.9494|±  |0.0143|\n|  - international_law                  |      1|none  |     0|acc   |↑  |0.9339|±  |0.0227|\n|  - jurisprudence                      |      1|none  |     0|acc   |↑  |0.9352|±  |0.0238|\n|  - logical_fallacies                  |      1|none  |     0|acc   |↑  |0.9080|±  |0.0227|\n|  - moral_disputes                     |      1|none  |     0|acc   |↑  |0.8468|±  |0.0194|\n|  - moral_scenarios                    |      1|none  |     0|acc   |↑  |0.5430|±  |0.0167|\n|  - philosophy                         |      1|none  |     0|acc   |↑  |0.8875|±  |0.0179|\n|  - prehistory                         |      1|none  |     0|acc   |↑  |0.8981|±  |0.0168|\n|  - professional_law                   |      1|none  |     0|acc   |↑  |0.6851|±  |0.0119|\n|  - world_religions                    |      1|none  |     0|acc   |↑  |0.9181|±  |0.0210|\n| - other                               |      2|none  |      |acc   |↑  |0.8680|±  |0.0058|\n|  - business_ethics                    |      1|none  |     0|acc   |↑  |0.8200|±  |0.0386|\n|  - clinical_knowledge                 |      1|none  |     0|acc   |↑  |0.8981|±  |0.0186|\n|  - college_medicine                   |      1|none  |     0|acc   |↑  |0.8671|±  |0.0259|\n|  - global_facts                       |      1|none  |     0|acc   |↑  |0.5900|±  |0.0494|\n|  - human_aging                        |      1|none  |     0|acc   |↑  |0.8610|±  |0.0232|\n|  - management                         |      1|none  |     0|acc   |↑  |0.8932|±  |0.0306|\n|  - marketing                          |      1|none  |     0|acc   |↑  |0.9487|±  |0.0145|\n|  - medical_genetics                   |      1|none  |     0|acc   |↑  |0.9700|±  |0.0171|\n|  - miscellaneous                      |      1|none  |     0|acc   |↑  |0.9272|±  |0.0093|\n|  - nutrition                          |      1|none  |     0|acc   |↑  |0.8954|±  |0.0175|\n|  - professional_accounting            |      1|none  |     0|acc   |↑  |0.7660|±  |0.0253|\n|  - professional_medicine              |      1|none  |     0|acc   |↑  |0.9412|±  |0.0143|\n|  - virology                           |      1|none  |     0|acc   |↑  |0.5602|±  |0.0386|\n| - social sciences                     |      2|none  |      |acc   |↑  |0.9123|±  |0.0050|\n|  - econometrics                       |      1|none  |     0|acc   |↑  |0.7895|±  |0.0384|\n|  - high_school_geography              |      1|none  |     0|acc   |↑  |0.9444|±  |0.0163|\n|  - high_school_government_and_politics|      1|none  |     0|acc   |↑  |0.9948|±  |0.0052|\n|  - high_school_macroeconomics         |      1|none  |     0|acc   |↑  |0.9333|±  |0.0126|\n|  - high_school_microeconomics         |      1|none  |     0|acc   |↑  |0.9664|±  |0.0117|\n|  - high_school_psychology             |      1|none  |     0|acc   |↑  |0.9633|±  |0.0081|\n|  - human_sexuality                    |      1|none  |     0|acc   |↑  |0.9160|±  |0.0243|\n|  - professional_psychology            |      1|none  |     0|acc   |↑  |0.8807|±  |0.0131|\n|  - public_relations                   |      1|none  |     0|acc   |↑  |0.7455|±  |0.0417|\n|  - security_studies                   |      1|none  |     0|acc   |↑  |0.8000|±  |0.0256|\n|  - sociology                          |      1|none  |     0|acc   |↑  |0.9403|±  |0.0168|\n|  - us_foreign_policy                  |      1|none  |     0|acc   |↑  |0.9300|±  |0.0256|\n| - stem                                |      2|none  |      |acc   |↑  |0.8468|±  |0.0062|\n|  - abstract_algebra                   |      1|none  |     0|acc   |↑  |0.7200|±  |0.0451|\n|  - anatomy                            |      1|none  |     0|acc   |↑  |0.8519|±  |0.0307|\n|  - astronomy                          |      1|none  |     0|acc   |↑  |0.9539|±  |0.0171|\n|  - college_biology                    |      1|none  |     0|acc   |↑  |0.9722|±  |0.0137|\n|  - college_chemistry                  |      1|none  |     0|acc   |↑  |0.6800|±  |0.0469|\n|  - college_computer_science           |      1|none  |     0|acc   |↑  |0.7900|±  |0.0409|\n|  - college_mathematics                |      1|none  |     0|acc   |↑  |0.6900|±  |0.0465|\n|  - college_physics                    |      1|none  |     0|acc   |↑  |0.7941|±  |0.0402|\n|  - computer_security                  |      1|none  |     0|acc   |↑  |0.8500|±  |0.0359|\n|  - conceptual_physics                 |      1|none  |     0|acc   |↑  |0.9234|±  |0.0174|\n|  - electrical_engineering             |      1|none  |     0|acc   |↑  |0.8345|±  |0.0310|\n|  - elementary_mathematics             |      1|none  |     0|acc   |↑  |0.8915|±  |0.0160|\n|  - high_school_biology                |      1|none  |     0|acc   |↑  |0.9613|±  |0.0110|\n|  - high_school_chemistry              |      1|none  |     0|acc   |↑  |0.8621|±  |0.0243|\n|  - high_school_computer_science       |      1|none  |     0|acc   |↑  |0.9400|±  |0.0239|\n|  - high_school_mathematics            |      1|none  |     0|acc   |↑  |0.6593|±  |0.0289|\n|  - high_school_physics                |      1|none  |     0|acc   |↑  |0.8278|±  |0.0308|\n|  - high_school_statistics             |      1|none  |     0|acc   |↑  |0.8704|±  |0.0229|\n|  - machine_learning                   |      1|none  |     0|acc   |↑  |0.7411|±  |0.0416|\n\n|      Groups      |Version|Filter|n-shot|Metric|   |Value |   |Stderr|\n|------------------|------:|------|------|------|---|-----:|---|-----:|\n|mmlu              |      2|none  |      |acc   |↑  |0.8361|±  |0.0029|\n| - humanities     |      2|none  |      |acc   |↑  |0.7579|±  |0.0059|\n| - other          |      2|none  |      |acc   |↑  |0.8680|±  |0.0058|\n| - social sciences|      2|none  |      |acc   |↑  |0.9123|±  |0.0050|\n| - stem           |      2|none  |      |acc   |↑  |0.8468|±  |0.0062|\n\nMMLU - Massive Multitask Language Understanding, ~14,000 multiple-choice questions across 57 subjects (math, history, law, medicine, etc.).\n\n-----\n\n## Quantizations\n\n| Filename | Quant | Description |\n|----------|-------|-------------|\n| Omega-Evolution-27B-v2.0-uncensored-heretic-BF16.gguf | BF16 | Full precision |\n| Omega-Evolution-27B-v2.0-uncensored-heretic-Q8_0.gguf | Q8_0 | Near-lossless, recommended |\n| Omega-Evolution-27B-v2.0-uncensored-heretic-Q6_K.gguf | Q6_K | Excellent quality |\n| Omega-Evolution-27B-v2.0-uncensored-heretic-Q5_K_M.gguf | Q5_K_M | Good balance |\n| Omega-Evolution-27B-v2.0-uncensored-heretic-Q5_K_S.gguf | Q5_K_S | Smaller Q5 |\n| Omega-Evolution-27B-v2.0-uncensored-heretic-Q4_K_M.gguf | Q4_K_M | Good for limited VRAM |\n\n## Vision Projector\n\n| Filename | Quant | Description |\n|----------|-------|-------------|\n| Omega-Evolution-27B-v2.0-mmproj-BF16.gguf | BF16 | Native precision |\n\nA Vision Projector File is Required for vision/multimodal capabilities. Use alongside any quantization above.\n\n## Usage\n\nWorks with llama.cpp, LM Studio, Ollama, and other GGUF-compatible tools.\n\n-----\n\n<style>\n.gs {\n  --bg:      #050a12;\n  --surface: #0a1019;\n  --edge:    #121c2c;\n  --rule:    #192638;\n  --text:    #8a9db4;\n  --dim:     #4e6880;\n  --bright:  #dce8f4;\n  --azure:   #0077cc;\n  --crimson: #00bbff;\n  --az-glow: rgba(0,119,204,0.10);\n  --cr-glow: rgba(0,187,255,0.06);\n  --mono:    'JetBrains Mono', monospace;\n  --sans:    'Inter', sans-serif;\n\n  font-family: var(--sans);\n  color: var(--text);\n  max-width: 900px;\n  margin: 0 auto;\n  padding: 0 0 60px;\n  line-height: 1.7;\n  font-size: 1rem;\n  background:\n    radial-gradient(ellipse at 50% 0%, rgba(0,140,255,0.04) 0%, transparent 50%),\n    radial-gradient(ellipse at 50% 100%, rgba(0,100,200,0.02) 0%, transparent 50%),\n    var(--bg);\n}\n\n/* ── Hero ── */\n.gs-hero {\n  position: relative;\n  overflow: hidden;\n}\n.gs-hero img {\n  display: block;\n  width: 100%;\n}\n.gs-ident {\n  position: absolute;\n  bottom: 0;\n  left: 0;\n  right: 0;\n  padding: 130px 48px 36px;\n  background: linear-gradient(\n    to top,\n    var(--bg) 0%,\n    rgba(5,10,18,0.92) 35%,\n    rgba(5,10,18,0.4) 65%,\n    transparent 100%\n  );\n}\n.gs-name {\n  font-family: var(--sans);\n  font-size: 3.2rem;\n  font-weight: 900;\n  color: var(--bright);\n  letter-spacing: 0.06em;\n  text-transform: uppercase;\n  line-height: 1;\n  margin: 0 0 10px;\n  text-shadow: 0 1px 2px rgba(0,0,0,0.6);\n}\n.gs-base {\n  font-family: var(--mono);\n  font-size: 0.68rem;\n  color: var(--crimson);\n  letter-spacing: 0.14em;\n  text-transform: uppercase;\n  display: block;\n}\n\n/* ── Sections ── */\n.gs-section {\n  padding: 0;\n}\n.gs-shead {\n  display: flex;\n  align-items: baseline;\n  gap: 14px;\n  padding: 16px 44px 14px;\n  margin-bottom: 28px;\n  border-top: 2px solid;\n  border-image: linear-gradient(90deg, var(--crimson), var(--azure)) 1;\n}\n.gs-snum {\n  font-family: var(--mono);\n  font-size: 0.68rem;\n  font-weight: 700;\n  color: var(--crimson);\n  letter-spacing: 0.12em;\n  flex-shrink: 0;\n}\n.gs-stitle {\n  font-size: 1.05rem;\n  font-weight: 700;\n  letter-spacing: 0.1em;\n  text-transform: uppercase;\n  color: var(--bright);\n}\n.gs-sbody {\n  padding: 0 44px 44px;\n}\n.gs-sbody p {\n  margin: 0 0 14px;\n  font-size: 0.95rem;\n}\n.gs-sbody p:last-child { margin-bottom: 0; }\n\n/* ── Data panels ── */\n.gs-stack {\n  display: flex;\n  flex-direction: column;\n  gap: 16px;\n}\n.gs-panel {\n  border: 1px solid var(--edge);\n  border-left: 2px solid var(--crimson);\n  position: relative;\n  background: var(--surface);\n  box-shadow: 0 0 20px rgba(0,140,255,0.03);\n}\n.gs-panel::before {\n  content: '';\n  position: absolute;\n  top: -1px;\n  right: -1px;\n  width: 10px;\n  height: 10px;\n  border-top: 1px solid var(--crimson);\n  border-right: 1px solid var(--crimson);\n  opacity: 0.5;\n}\n.gs-panel::after {\n  content: '';\n  position: absolute;\n  bottom: -1px;\n  right: -1px;\n  width: 10px;\n  height: 10px;\n  border-bottom: 1px solid var(--azure);\n  border-right: 1px solid var(--azure);\n  opacity: 0.4;\n}\n.gs-panel-head {\n  font-family: var(--mono);\n  font-size: 0.68rem;\n  font-weight: 700;\n  letter-spacing: 0.14em;\n  text-transform: uppercase;\n  color: var(--dim);\n  padding: 10px 16px;\n  border-bottom: 1px solid var(--edge);\n}\n.gs-row {\n  display: grid;\n  grid-template-columns: 10ch 1fr;\n  align-items: baseline;\n  column-gap: 4px;\n  padding: 9px 16px;\n  border-bottom: 1px solid var(--edge);\n  font-size: 0.9rem;\n}\n.gs-row:last-child { border-bottom: none; }\n.gs-key {\n  font-family: var(--mono);\n  font-size: 0.9rem;\n  color: var(--dim);\n}\n.gs-key::after {\n  content: ':';\n}\n.gs-val {\n  color: var(--bright);\n  font-size: 0.9rem;\n}\n.gs-row .gs-val:only-child {\n  grid-column: 1 / -1;\n}\n\n/* ── Quantizations ── */\n.gs-qrow {\n  display: flex;\n  gap: 12px;\n  flex-wrap: wrap;\n}\n.gs-qpanel {\n  background: var(--surface);\n  border: 1px solid var(--edge);\n  border-left: 2px solid var(--crimson);\n  display: flex;\n  align-items: center;\n  gap: 16px;\n  padding: 12px 18px;\n  position: relative;\n  box-shadow: 0 0 20px rgba(0,140,255,0.03);\n}\n.gs-qpanel::before {\n  content: '';\n  position: absolute;\n  top: -1px;\n  right: -1px;\n  width: 10px;\n  height: 10px;\n  border-top: 1px solid var(--crimson);\n  border-right: 1px solid var(--crimson);\n  opacity: 0.5;\n}\n.gs-qpanel::after {\n  content: '';\n  position: absolute;\n  bottom: -1px;\n  right: -1px;\n  width: 10px;\n  height: 10px;\n  border-bottom: 1px solid var(--azure);\n  border-right: 1px solid var(--azure);\n  opacity: 0.4;\n}\n.gs-qtype {\n  font-family: var(--mono);\n  font-size: 0.58rem;\n  font-weight: 700;\n  letter-spacing: 0.18em;\n  text-transform: uppercase;\n  color: var(--crimson);\n  flex-shrink: 0;\n}\n.gs-qsep {\n  width: 1px;\n  height: 16px;\n  background: var(--rule);\n  flex-shrink: 0;\n}\n.gs-qpanel a {\n  color: var(--bright);\n  text-decoration: none;\n  font-size: 0.9rem;\n  border-bottom: 1px solid var(--rule);\n}\n.gs-qpanel a:hover { color: var(--crimson); border-bottom-color: var(--crimson); }\n\n/* ── Links ── */\n.gs a {\n  color: var(--bright);\n  text-decoration: none;\n  border-bottom: 1px solid var(--rule);\n}\n.gs a:hover { color: var(--crimson); border-bottom-color: var(--crimson); }\n\n/* ── Dropdown ── */\n.gs details {\n  border: 1px solid var(--edge);\n  border-left: 2px solid var(--crimson);\n  margin-top: 24px;\n  position: relative;\n  background: var(--surface);\n  box-shadow: 0 0 20px rgba(0,140,255,0.03);\n}\n.gs details::before {\n  content: '';\n  position: absolute;\n  top: -1px;\n  right: -1px;\n  width: 10px;\n  height: 10px;\n  border-top: 1px solid var(--crimson);\n  border-right: 1px solid var(--crimson);\n  opacity: 0.5;\n}\n.gs details::after {\n  content: '';\n  position: absolute;\n  bottom: -1px;\n  right: -1px;\n  width: 10px;\n  height: 10px;\n  border-bottom: 1px solid var(--azure);\n  border-right: 1px solid var(--azure);\n  opacity: 0.4;\n}\n.gs summary {\n  list-style: none;\n  padding: 11px 16px;\n  cursor: pointer;\n  font-family: var(--mono);\n  font-size: 0.72rem;\n  font-weight: 700;\n  letter-spacing: 0.12em;\n  text-transform: uppercase;\n  color: var(--dim);\n  user-select: none;\n  display: flex;\n  align-items: center;\n  gap: 10px;\n}\n.gs summary::-webkit-details-marker { display: none; }\n.gs summary::before {\n  content: '+';\n  color: var(--crimson);\n  font-size: 1rem;\n  line-height: 1;\n  flex-shrink: 0;\n}\n.gs details[open] summary::before { content: '−'; }\n.gs summary:hover { color: var(--bright); }\n.gs-detail-body {\n  padding: 22px 18px;\n  border-top: 1px solid var(--edge);\n}\n.gs-detail-body p { margin: 0 0 16px; font-size: 0.9rem; }\n.gs-cfg-title {\n  font-family: var(--mono);\n  font-size: 0.72rem;\n  font-weight: 700;\n  letter-spacing: 0.1em;\n  text-transform: uppercase;\n  color: var(--dim);\n  margin: 0 0 8px;\n}\n\n/* ── Code ── */\n.gs pre {\n  background: #030810;\n  border: 1px solid var(--edge);\n  border-left: 2px solid var(--azure);\n  padding: 16px 18px;\n  overflow-x: auto;\n  font-family: var(--mono);\n  font-size: 0.76rem;\n  line-height: 1.6;\n  color: var(--text);\n  margin: 0 0 22px;\n}\n.gs pre:last-child { margin-bottom: 0; }\n.gs pre code { background: none; color: inherit; padding: 0; }\n.gs code {\n  font-family: var(--mono);\n  font-size: 0.875em;\n  color: var(--crimson);\n  background: var(--az-glow);\n  padding: 2px 5px;\n}\n</style>\n<html lang=\"en\">\n<head>\n<meta charset=\"UTF-8\">\n<meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n<title>BlueStar</title>\n<link rel=\"preconnect\" href=\"https://fonts.googleapis.com\">\n<link rel=\"preconnect\" href=\"https://fonts.gstatic.com\" crossorigin>\n<link href=\"https://fonts.googleapis.com/css2?family=Inter:wght@400;600;700;900&family=JetBrains+Mono:wght@400;700&display=swap\" rel=\"stylesheet\">\n</head>\n<body>\n<div class=\"gs\">\n\n<div class=\"gs-hero\">\n<img src=\"https://cdn-uploads.huggingface.co/production/uploads/65b19c6c638328850e12d38c/ruIMSzVP1DU1_HxUqBmFx.png\" alt=\"image\">\n<div class=\"gs-ident\">\n<h1 class=\"gs-name\">BlueStar v2</h1>\n<span class=\"gs-base\">Qwen3.5 27B</span>\n</div>\n</div>\n\n\n<div class=\"gs-section\">\n<div class=\"gs-shead\">\n<span class=\"gs-snum\">01</span>\n<span class=\"gs-stitle\">Overview</span>\n</div>\n<div class=\"gs-sbody\">\n<p>Designed for RP and writing tasks.</p>\n<p>Feels like a good improvement on v1. This version aims to fix the rep and improve the intelligence while keeping the creativity.</p>\n<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>\n</div>\n</div>\n\n\n<div class=\"gs-section\">\n<div class=\"gs-shead\">\n<span class=\"gs-snum\">02</span>\n<span class=\"gs-stitle\">SillyTavern Settings</span>\n</div>\n<div class=\"gs-sbody\">\n<div class=\"gs-stack\">\n<div class=\"gs-panel\">\n<div class=\"gs-panel-head\">Recommended Roleplay Format</div>\n<div class=\"gs-row\"><span class=\"gs-key\">Actions</span><span class=\"gs-val\">In plaintext</span></div>\n<div class=\"gs-row\"><span class=\"gs-key\">Dialogue</span><span class=\"gs-val\">\"In quotes\"</span></div>\n<div class=\"gs-row\"><span class=\"gs-key\">Thoughts</span><span class=\"gs-val\">*In asterisks*</span></div>\n</div>\n<div class=\"gs-panel\">\n<div class=\"gs-panel-head\">Recommended Samplers</div>\n<div class=\"gs-row\"><span class=\"gs-key\">Temp</span><span class=\"gs-val\">0.8 - 1.0</span></div>\n<div class=\"gs-row\"><span class=\"gs-key\">MinP</span><span class=\"gs-val\">0.05 - 0.075</span></div>\n</div>\n<div class=\"gs-panel\">\n<div class=\"gs-panel-head\">Instruct</div>\n<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>\n<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>\n</div>\n</div>\n</div>\n</div>\n\n\n<div class=\"gs-section\">\n<div class=\"gs-shead\">\n<span class=\"gs-snum\">03</span>\n<span class=\"gs-stitle\">Quantizations</span>\n</div>\n<div class=\"gs-sbody\">\n<div class=\"gs-qrow\">\n<div class=\"gs-qpanel\">\n<span class=\"gs-qtype\">GGUF</span>\n<div class=\"gs-qsep\"></div>\n<a href=\"https://huggingface.co/zerofata/Q3.5-BlueStar-v2-27B-GGUF\">iMatrix</a>\n</div>\n</div>\n</div>\n</div>\n\n\n<div class=\"gs-section\">\n<div class=\"gs-shead\">\n<span class=\"gs-snum\">04</span>\n<span class=\"gs-stitle\">Creation Process</span>\n</div>\n<div class=\"gs-sbody\">\n<p>Creation Process: SFT</p>\n<p>SFT on approx 27 million tokens.</p>\n<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>\n<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>\n<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>\n<p>Trained using Axolotl.</p>\n<details>\n<summary>Axolotl Config</summary>\n<div class=\"gs-detail-body\">\n<div class=\"gs-cfg-title\">SFT (4&#215;H200)</div>\n<pre><code>base_model: Qwen/Qwen3.5&#45;27B\n&#32;\nplugins:\n  &#45; axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin\nstrict: false\n&#32;\ndatasets:\n  &#45; path: ./data/bluestar_v2_sft_3_all_rp_attempt_masked_20260318_075236.jsonl\n&#32;\nval_set_size: 0.02\noutput_dir: ./Qwen3.5&#45;27B&#45;v2&#45;SFT&#45;5\n&#32;\nsequence_len: 10756\nsample_packing: true\n&#32;\nload_in_8bit: true\nadapter: lora\nlora_r: 128\nlora_alpha: 128\npeft_use_rslora: true\nlora_target_modules:\n  &#45; q_proj\n  &#45; k_proj\n  &#45; v_proj\n  &#45; o_proj\n  &#45; down_proj\n  &#45; up_proj\n  &#35; Uncomment below to also target the linear attention projections.\n  &#35; These use separate in_proj_qkv / in_proj_z / out_proj (Qwen3.5&#45;specific).\n  &#45; linear_attn.in_proj_qkv\n  &#45; linear_attn.in_proj_z\n  &#45; linear_attn.out_proj\n&#32;\nwandb_project: Qwen3.5&#45;27B&#45;SFT\nwandb_name: Qwen3.5&#45;27B&#45;v2&#45;SFT&#45;5\n&#32;\ngradient_accumulation_steps: 4\nmicro_batch_size: 1\nnum_epochs: 2\noptimizer: adamw_torch_8bit\nlr_scheduler: cosine\nlearning_rate: 1.2e&#45;5\nweight_decay: 0.01\nwarmup_ratio: 0.05\n&#32;\nbf16: auto\ntf32: true\n&#32;\nresume_from_checkpoint:\nlogging_steps: 1\nflash_attention: true\n&#32;\nevals_per_epoch: 4\nsaves_per_epoch: 4\nspecial_tokens:\n&#32;\nfsdp_config:\n  fsdp_version: 2\n  offload_params: false\n  cpu_ram_efficient_loading: false\n  auto_wrap_policy: TRANSFORMER_BASED_WRAP\n  transformer_layer_cls_to_wrap: Qwen3_5DecoderLayer\n  state_dict_type: FULL_STATE_DICT\n  sharding_strategy: FULL_SHARD\n  reshard_after_forward: true\n  activation_checkpointing: true</code></pre>\n</div>\n</details>\n</div>\n</div>\n\n</div>\n</body>\n</html>",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "nsfw",
    "explicit",
    "roleplay",
    "unaligned",
    "dangerous",
    "ERP",
    "heretic",
    "uncensored",
    "decensored",
    "abliterated",
    "ara",
    "en",
    "base_model:llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic",
    "base_model:quantized:llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic",
    "license:apache-2.0",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 895,
  "gated": false,
  "private": false,
  "last_modified": "2026-03-27T19:48:33.000Z",
  "created_at": "2026-03-26T23:52:08.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "69c5c6a8830d81469455e361",
  "id": "llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic-GGUF",
  "modelId": "llmfan46/Omega-Evolution-27B-v2.0-uncensored-heretic-GGUF",
  "sha": "a2b35f7cb2c099cc5920b79c50b8051805938861",
  "createdAt": "2026-03-26T23:52:08.000Z",
  "lastModified": "2026-03-27T19:48:33.000Z",
  "author": "llmfan46",
  "downloads": 895,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 9
}