Gliese-Qwen3.5-0.8B-Abliterated-Caption

Gliese-Qwen3.5-0.8B-Abliterated-Caption

Gliese-Qwen3.5-0.8B-Abliterated-Caption is an abliterated evolution built on top of Qwen/Qwen3.5-0.8B, designed specifically for generalized and unfiltered image captioning. The model applies advanced refusal direction analysis and abliterated training strategies to minimize internal refusal behaviors while maximizing descriptive capability and visual understanding. The result is a lightweight 0.8B parameter vision-language model optimized for detailed captions, strong scene understanding, and rich visual descriptions, while remaining highly efficient.

[!IMPORTANT] This model is intended for research and learning purposes only. The model has reduced internal refusal behaviors, and any content generated by it is used at the user’s own risk. The authors and hosting page disclaim any liability for content generated by this model. Users are responsible for ensuring that the model is used in a a safe, ethical, and lawful manner.

<div style=" background: rgba(255, 235, 59, 0.15); padding: 16px; border-radius: 6px; border: 1px solid rgba(255, 235, 59, 0.4); margin: 16px 0; "> <details> <summary>Get GGUF</summary> </details> </div>

[!NOTE] Expert Image Captioning System (chat_template.jinja) – https://huggingface.co/prithivMLmods/Gliese-Qwen3.5-0.8B-Abliterated-Caption/blob/main/chat_template.jinja [Recommended]

[!NOTE] Standard or Default (chat_template.jinja) – https://huggingface.co/prithivMLmods/Gliese-Qwen3.5-0.8B-Abliterated-Caption/blob/main/standard-chat_template/chat_template.jinja

---

Base Model Signatures:

This model has been re-sharded and optimized for the latest Transformers version from the base model: https://huggingface.co/huihui-ai/Huihui-Qwen3.5-0.8B-abliterated.

---

Download the model

hf auth login --token <YOUR_HF_TOKEN>

hf download prithivMLmods/Gliese-Qwen3.5-0.8B-Abliterated-Caption

Key Highlights

• Advanced Refusal Direction Analysis Uses targeted activation analysis to identify and mitigate refusal directions within the model’s latent space.

• Abliterated Caption Training Fine-tuned for unfiltered and detailed caption generation, enabling comprehensive visual descriptions without excessive refusal behaviors.

• Optimized Visual Understanding Enhanced to provide rich, context-aware descriptions of scenes, objects, people, and environments.

• 0.8B Parameter Architecture Built on Qwen3.5-0.8B, delivering efficient multimodal reasoning and caption generation while remaining lightweight and fast to deploy.

• High-Fidelity Caption Generation Designed to produce structured and semantically detailed captions suitable for dataset generation, annotation, and research.

• Efficient Deployment Suitable for caption dataset creation, multimodal research, local inference pipelines, edge deployments, and AI development workflows.

Quick Start with Transformers

pip install transformers==5.3.0
# or
pip install git+https://github.com/huggingface/transformers.git

from transformers import Qwen3_5ForConditionalGeneration, AutoProcessor
import torch

model = Qwen3_5ForConditionalGeneration.from_pretrained(
    "prithivMLmods/Gliese-Qwen3.5-0.8B-Abliterated-Caption",
    torch_dtype="auto",
    device_map="auto"
)

processor = AutoProcessor.from_pretrained(
    "prithivMLmods/Gliese-Qwen3.5-0.8B-Abliterated-Caption"
)

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Describe this image in extreme detail."}
        ],
    }
]

text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)

inputs = processor(
    text=[text],
    padding=True,
    return_tensors="pt"
).to("cuda")

generated_ids = model.generate(**inputs, max_new_tokens=512)

generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]

output_text = processor.batch_decode(
    generated_ids_trimmed,
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False
)

print(output_text)

Intended Use

• High-Detail Image Captioning – Generating descriptive captions for images.
• Dataset Generation – Creating large-scale caption datasets for multimodal training.
• Vision-Language Research – Studying multimodal reasoning and captioning behavior.
• Annotation Automation – Assisting in automatic labeling and visual description tasks.
• Lightweight Multimodal Deployment – Running captioning models efficiently on local machines and edge devices.

Limitations & Risks

Important Note: This model intentionally reduces built-in refusal mechanisms.

• Unfiltered Outputs – The model may generate explicit or controversial captions depending on the input images.
• User Responsibility – Generated outputs should be handled responsibly and within legal and ethical boundaries.
• Model Capacity Limits – Due to its 0.8B parameter size, the model may have reduced reasoning depth compared to larger multimodal architectures.