license: apache-2.0 base_model:

• Qwen/Qwen3.5-4B language:
• en pipeline_tag: video-text-to-text library_name: transformers tags:
• text-generation-inference
• reinforcement-learning
• video-text-to-text
• alignment-training
• RLHF
• RFT
• video-understanding
• video-classification
• video-safety
• content-safety
• content-moderation
• safety-classifier
• guardrail datasets:
• prithivMLmods/OpenVideo-Scene-Reasoning
• PALM-Lab/vid-guard-rlhf-unsafe

So, C3 is definitely relevant because a weapon is present. C5 might be relevant if we consider guns as regulated activity. The prompt says "C5 covers... weapons trafficking". A gun on a table might be considered a weapon. I will include C3 and maybe C5 if it fits well. The prompt asks for "which category codes apply, if any, and briefly why". The image is "safe or not" - this is the user's question. I need to answer that in the output.

Line 1: DESCRIPTION: A young boy peeks over a table with a handgun resting on the surface in front of him. Line 2: EXPLANATION: C3 (Threats, Violence & Harm) is present due to the handgun on the table. C5 (Illegal/Regulated Activities) is relevant as a firearm is a regulated item. Line 3: GUARDRAIL: {"C1(Sexual Content)": false, "C2(Harassment & Bullying)": false, "C3(Threats, Violence & Harm)": true, "C4(False & Deceptive Information)": false, "C5(Illegal/Regulated Activities)": true, "C6(Hateful Content & Extremism)": false}

Wait, the prompt says "C5 covers illegal drugs, contraband, weapons trafficking, or other clearly regulated or unlawful activity." A gun is a weapon. So C5 is appropriate. The prompt says "C3 covers weapons pointed at people, physical violence, threats of harm, self-harm, or injury to people or animals." The gun is on a table, not pointed at a person, but it is a weapon. It's a bit ambiguous. But since it's a weapon, C3 is a good fit. The prompt says "Set a key to true only when the content clearly and directly falls into that category, and false otherwise; do not default to true for merely adjacent or ambiguous themes." The gun is clearly a weapon. So C3 is true. The gun is clearly a regulated activity (firearms). So C5 is true.

Let's refine the description. Line 1: DESCRIPTION: A young boy peeks over a table with a handgun resting on the surface in front of him. Line 2: EXPLANATION: C3 (Threats, Violence & Harm) is present due to the handgun on the table. C5 (Illegal/Regulated Activities) is relevant as a firearm is a regulated item. Line 3: GUARDRAIL: {"C1(Sexual Content)": false, "C2(Harassment & Bullying)": false, "C3(Threats, Violence & Harm)": true, "C4(False & Deceptive Information)": false, "C5(Illegal/Regulated Activities)": true, "C6(Hateful Content & Extremism)": false}

Double check the order of keys in GUARDRAIL. "C1(Sexual Content)", "C2(Harassment & Bullying)", "C3(Threats, Violence & Harm)", "C4(False & Deceptive Information)", "C5(Illegal/Regulated Activities)", "C6(Hateful Content & Extremism)" Yes, that matches the prompt.

One more check on the description. "A young boy peeks over a table with a handgun resting on the surface in front of him." - good. One more check on the explanation. "C3 (Threats, Violence & Harm) is present due to the handgun on the table. C5 (Illegal/Regulated Activities) is relevant as a firearm is a regulated item." - good. One more check on the guardrail. {"C1(Sexual Content)": false, "C2(Harassment & Bullying)": false, "C3(Threats, Violence & Harm)": true, "C4(False & Deceptive Information)": false, "C5(Illegal/Regulated Activities)": true, "C6(Hateful Content & Extremism)": false} - good.

Final check on the output format.

• Three lines.
• Line one begins with "DESCRIPTION: "
• Line two begins with "EXPLANATION: "
• Line three begins with "GUARDRAIL: "
• No markdown formatting.
• No text outside the three lines.

Looks good.

VideoGuard-Qwen3.5-4B-Safety-RL-Uncensored

VideoGuard-Qwen3.5-4B-Safety-RL-Uncensored is a multimodal safety classifier built on top of Qwen/Qwen3.5-4B. The model was trained on a mixture of approximately 10,000 video safety and scene-reasoning samples to analyze video content and classify potentially unsafe content across predefined safety categories. The model is designed to generate a structured DESCRIPTION, EXPLANATION, and GUARDRAIL output, making it suitable for video content filtering, safety evaluation, and multimodal guardrail research.

[!NOTE] This model is an experimental release and may generate unexpected classifications or reasoning artifacts in certain scenarios. Safety classifications should be treated as model predictions rather than definitive judgments.

Key Highlights

• Qwen 3.5 Multimodal Backbone: Built on top of Qwen/Qwen3.5-4B.
• Video Safety Classification: Designed to analyze video content and identify potentially unsafe or sensitive material.
• 10K Training Samples: Trained using a mixture of approximately 10,000 video safety and scene-reasoning samples.
• Structured Guardrail Output: Produces a description, explanation, and structured C1–C6 safety classification.
• Multimodal Reasoning: Uses visual and textual information to analyze video scenes and determine applicable safety categories.
• Safety Evaluation: Designed for content filtering, safety evaluation, red teaming, and multimodal guardrail research.

Safety Categories

The model classifies content across six predefined categories:

Output Format

The model is trained to produce three primary components:

DESCRIPTION: <summary of the video content>

EXPLANATION: <reasoning explaining the classification>

GUARDRAIL: {
  "C1(Sexual Content)": false,
  "C2(Harassment & Bullying)": false,
  "C3(Threats, Violence & Harm)": false,
  "C4(False & Deceptive Information)": false,
  "C5(Illegal/Regulated Activities)": false,
  "C6(Hateful Content & Extremism)": false
}

Example — Safe Content

DESCRIPTION: A hand holds four small, leaf-decorated candies in green, orange, and teal colors against a wooden table background.

EXPLANATION: The content is ordinary and safe, depicting a simple everyday scene of someone holding candy. It does not contain any of the specific category codes (C1-C6) as defined in the prompt.

GUARDRAIL: {
  "C1(Sexual Content)": false,
  "C2(Harassment & Bullying)": false,
  "C3(Threats, Violence & Harm)": false,
  "C4(False & Deceptive Information)": false,
  "C5(Illegal/Regulated Activities)": false,
  "C6(Hateful Content & Extremism)": false
}

Example — C1 Classification

DESCRIPTION: A woman with blonde hair tied in a ponytail poses against a plain white wall, wearing a light blue blazer and trousers with an open front that exposes her midsection.

EXPLANATION: This content is classified under C1 based on the visual presentation described in the input.

GUARDRAIL: {
  "C1(Sexual Content)": true,
  "C2(Harassment & Bullying)": false,
  "C3(Threats, Violence & Harm)": false,
  "C4(False & Deceptive Information)": false,
  "C5(Illegal/Regulated Activities)": false,
  "C6(Hateful Content & Extremism)": false
}

Quick Start with Transformers

pip install transformers
pip install accelerate

from transformers import Qwen3_5ForConditionalGeneration, AutoProcessor
import torch

model = Qwen3_5ForConditionalGeneration.from_pretrained(
    "prithivMLmods/VideoGuard-Qwen3.5-4B-Safety-RL-Uncensored",
    torch_dtype="auto",
    device_map="auto"
)

processor = AutoProcessor.from_pretrained(
    "prithivMLmods/VideoGuard-Qwen3.5-4B-Safety-RL-Uncensored"
)

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "Analyze this video and classify it using the C1-C6 guardrail categories."
            }
        ],
    }
]

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=256
)

output_text = processor.batch_decode(
    [
        out[len(inp):]
        for inp, out in zip(inputs.input_ids, generated_ids)
    ],
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False
)

print(output_text[0])

Training Details

Training Datasets

• prithivMLmods/OpenVideo-Scene-Reasoning: Video scene reasoning dataset used for multimodal video understanding and scene analysis.

• PALM-Lab/vid-guard-rlhf-unsafe: Video safety dataset containing examples for safety classification and guardrail behavior.

Intended Use

• Video Content Filtering: Classifying potentially unsafe video content.
• Safety Evaluation: Evaluating multimodal safety behavior across predefined categories.
• Video Guardrails: Building automated safety-filtering pipelines for video applications.
• Red Teaming: Testing multimodal models against challenging safety scenarios.
• Multimodal Research: Studying video understanding and safety classification.
• Content Moderation: Supporting automated video moderation workflows.

Limitations

• Experimental Model: The model may produce incorrect or inconsistent classifications.
• False Positives: Benign content may occasionally be classified as unsafe.
• False Negatives: Unsafe content may occasionally be missed.
• Context Sensitivity: Classification accuracy can depend heavily on the available visual context and prompt.
• Model Predictions: C1–C6 classifications should be treated as model predictions and should not be considered definitive safety judgments.

Acknowledgements

• Qwen/Qwen3.5-4B: Base multimodal model used for this project.

• TRL – Transformers Reinforcement Learning: TRL is a full-stack library providing tools to train transformer language models with methods including Supervised Fine-Tuning (SFT), Group Relative Policy Optimization (GRPO), Direct Preference Optimization (DPO), Reward Modeling, and more.

• Transformers: Transformers provides state-of-the-art machine learning models for text, computer vision, audio, video, and multimodal tasks, supporting both inference and training.