DocShield-4B API & Inference Endpoint

Training data

DocShield-4B was developed using the RealText forensic document datasets:

Capabilities

Visual forgery trace analysis — crude redaction / mosaic, font & anti-aliasing inconsistency, edge halos, copy-paste artifacts, compression mismatches.
Logical & fact-checking — price/quantity contradictions, date conflicts, bulk-discount logic violations.
Semantic alteration detection — subtle spec substitutions (e.g. color, material) that bypass crude visual checks.
Localization — bounding-box coordinates for each tampered region with per-anomaly reasoning.
Structured report — conclusion (FORGED / authentic) + fraud-risk score.

Model details

Table

Base model	Qwen3.5-VL-4B
Architecture	Qwen3_5ForConditionalGeneration (hybrid linear / full attention)
Precision (weights)	bfloat16
Max new tokens	1024 (default)
Thinking mode	supported (`--thinking` / `--no-thinking`)

The base model is not bundled here. This repository only releases the fine-tuned DocShield-4B weights.

Quick start

Install dependencies:

bash

pip install -U transformers torch torchvision pillow qwen-vl-utils

Requires a transformers version with native Qwen3.5-VL (qwen3_5) support. The released weights were saved with transformers==5.13.0.

Run inference (see inference.py in this repo):

bash

# default: greedy, thinking disabledpython inference.py --image path/to/image.jpg --no-thinking --max-new-tokens 1024 # with sampling + thinking modepython inference.py --image path/to/image.jpg --thinking --do-sample \                    --temperature 0.6 --top-p 0.8 --top-k 20 --max-new-tokens 2048

Minimal example

python

import torchfrom transformers import AutoProcessor, AutoTokenizer, AutoModelForImageTextToTextfrom qwen_vl_utils import process_vision_info MODEL = "vankey/DocShield-4B" tokenizer = AutoTokenizer.from_pretrained(MODEL, trust_remote_code=True)processor = AutoProcessor.from_pretrained(MODEL, trust_remote_code=True)model = AutoModelForImageTextToText.from_pretrained(    MODEL, torch_dtype=torch.bfloat16, trust_remote_code=True, device_map="auto")model.eval() SYSTEM_PROMPT = (    "你是一个图像鉴伪专家，擅长结合视觉，文字结合伪造特征分析手段鉴别输入图像的真假。"    "分析过程中，你会逐步分析，抽丝剥茧，找到图像伪造的蛛丝马迹，最终给出专业的鉴别结果及分析。")USER_PROMPT = "请分析这张文档图片是否存在伪造或篡改风险，并输出一份专业、精炼、准确的防伪分析报告。" messages = [    {"role": "system", "content": [{"type": "text", "text": SYSTEM_PROMPT}]},    {"role": "user", "content": [        {"type": "image", "image": "image.jpg"},        {"type": "text", "text": USER_PROMPT},    ]},] text = processor.apply_chat_template(messages, tokenize=False,                                     add_generation_prompt=True, enable_thinking=False)image_inputs, video_inputs = process_vision_info(messages)inputs = processor(text=[text], images=image_inputs, videos=video_inputs,                   padding=True, return_tensors="pt").to(model.device) with torch.no_grad():    out = model.generate(**inputs, max_new_tokens=1024, do_sample=False) gen = [o[len(i):] for i, o in zip(inputs["input_ids"], out)]print(processor.batch_decode(gen, skip_special_tokens=False)[0])

Inference notes

Image input — pass the image path directly in the message content; the processor handles resize/tokenization.
Thinking mode — --no-thinking (default) gives a direct report; --thinking enables Qwen3.5 chain-of-thought before the report (use a larger --max-new-tokens).
Decoding — greedy by default (do_sample=False); pass --do-sample with --temperature/--top-p/--top-k for sampling.
Precision — bfloat16 is the tested configuration (--dtype bf16).

Citation

bibtex

@article{docshield2026,  title={DocShield: A Forensic Vision-Language Model for Document Forgery Analysis},  author={DocShield},  year={2026},  url={https://arxiv.org/abs/2604.02694}}

License

Apache-2.0.

Training data

DocShield-4B was developed using the RealText forensic document datasets:

Capabilities

Visual forgery trace analysis — crude redaction / mosaic, font & anti-aliasing inconsistency, edge halos, copy-paste artifacts, compression mismatches.
Logical & fact-checking — price/quantity contradictions, date conflicts, bulk-discount logic violations.
Semantic alteration detection — subtle spec substitutions (e.g. color, material) that bypass crude visual checks.
Localization — bounding-box coordinates for each tampered region with per-anomaly reasoning.
Structured report — conclusion (FORGED / authentic) + fraud-risk score.

Model details

Table

Base model	Qwen3.5-VL-4B
Architecture	Qwen3_5ForConditionalGeneration (hybrid linear / full attention)
Precision (weights)	bfloat16
Max new tokens	1024 (default)
Thinking mode	supported (`--thinking` / `--no-thinking`)

The base model is not bundled here. This repository only releases the fine-tuned DocShield-4B weights.

Quick start

Install dependencies:

bash

pip install -U transformers torch torchvision pillow qwen-vl-utils

pip install -U transformers torch torchvision pillow qwen-vl-utils

Requires a transformers version with native Qwen3.5-VL (qwen3_5) support. The released weights were saved with transformers==5.13.0.

Run inference (see inference.py in this repo):

bash

# default: greedy, thinking disabledpython inference.py --image path/to/image.jpg --no-thinking --max-new-tokens 1024 # with sampling + thinking modepython inference.py --image path/to/image.jpg --thinking --do-sample \                    --temperature 0.6 --top-p 0.8 --top-k 20 --max-new-tokens 2048

# default: greedy, thinking disabledpython inference.py --image path/to/image.jpg --no-thinking --max-new-tokens 1024 # with sampling + thinking modepython inference.py --image path/to/image.jpg --thinking --do-sample \                    --temperature 0.6 --top-p 0.8 --top-k 20 --max-new-tokens 2048

Minimal example

python

import torchfrom transformers import AutoProcessor, AutoTokenizer, AutoModelForImageTextToTextfrom qwen_vl_utils import process_vision_info MODEL = "vankey/DocShield-4B" tokenizer = AutoTokenizer.from_pretrained(MODEL, trust_remote_code=True)processor = AutoProcessor.from_pretrained(MODEL, trust_remote_code=True)model = AutoModelForImageTextToText.from_pretrained(    MODEL, torch_dtype=torch.bfloat16, trust_remote_code=True, device_map="auto")model.eval() SYSTEM_PROMPT = (    "你是一个图像鉴伪专家，擅长结合视觉，文字结合伪造特征分析手段鉴别输入图像的真假。"    "分析过程中，你会逐步分析，抽丝剥茧，找到图像伪造的蛛丝马迹，最终给出专业的鉴别结果及分析。")USER_PROMPT = "请分析这张文档图片是否存在伪造或篡改风险，并输出一份专业、精炼、准确的防伪分析报告。" messages = [    {"role": "system", "content": [{"type": "text", "text": SYSTEM_PROMPT}]},    {"role": "user", "content": [        {"type": "image", "image": "image.jpg"},        {"type": "text", "text": USER_PROMPT},    ]},] text = processor.apply_chat_template(messages, tokenize=False,                                     add_generation_prompt=True, enable_thinking=False)image_inputs, video_inputs = process_vision_info(messages)inputs = processor(text=[text], images=image_inputs, videos=video_inputs,                   padding=True, return_tensors="pt").to(model.device) with torch.no_grad():    out = model.generate(**inputs, max_new_tokens=1024, do_sample=False) gen = [o[len(i):] for i, o in zip(inputs["input_ids"], out)]print(processor.batch_decode(gen, skip_special_tokens=False)[0])

import torchfrom transformers import AutoProcessor, AutoTokenizer, AutoModelForImageTextToTextfrom qwen_vl_utils import process_vision_info MODEL = "vankey/DocShield-4B" tokenizer = AutoTokenizer.from_pretrained(MODEL, trust_remote_code=True)processor = AutoProcessor.from_pretrained(MODEL, trust_remote_code=True)model = AutoModelForImageTextToText.from_pretrained(    MODEL, torch_dtype=torch.bfloat16, trust_remote_code=True, device_map="auto")model.eval() SYSTEM_PROMPT = (    "你是一个图像鉴伪专家，擅长结合视觉，文字结合伪造特征分析手段鉴别输入图像的真假。"    "分析过程中，你会逐步分析，抽丝剥茧，找到图像伪造的蛛丝马迹，最终给出专业的鉴别结果及分析。")USER_PROMPT = "请分析这张文档图片是否存在伪造或篡改风险，并输出一份专业、精炼、准确的防伪分析报告。" messages = [    {"role": "system", "content": [{"type": "text", "text": SYSTEM_PROMPT}]},    {"role": "user", "content": [        {"type": "image", "image": "image.jpg"},        {"type": "text", "text": USER_PROMPT},    ]},] text = processor.apply_chat_template(messages, tokenize=False,                                     add_generation_prompt=True, enable_thinking=False)image_inputs, video_inputs = process_vision_info(messages)inputs = processor(text=[text], images=image_inputs, videos=video_inputs,                   padding=True, return_tensors="pt").to(model.device) with torch.no_grad():    out = model.generate(**inputs, max_new_tokens=1024, do_sample=False) gen = [o[len(i):] for i, o in zip(inputs["input_ids"], out)]print(processor.batch_decode(gen, skip_special_tokens=False)[0])

Inference notes

Image input — pass the image path directly in the message content; the processor handles resize/tokenization.
Thinking mode — --no-thinking (default) gives a direct report; --thinking enables Qwen3.5 chain-of-thought before the report (use a larger --max-new-tokens).
Decoding — greedy by default (do_sample=False); pass --do-sample with --temperature/--top-p/--top-k for sampling.
Precision — bfloat16 is the tested configuration (--dtype bf16).

Citation

bibtex

@article{docshield2026,  title={DocShield: A Forensic Vision-Language Model for Document Forgery Analysis},  author={DocShield},  year={2026},  url={https://arxiv.org/abs/2604.02694}}

@article{docshield2026,  title={DocShield: A Forensic Vision-Language Model for Document Forgery Analysis},  author={DocShield},  year={2026},  url={https://arxiv.org/abs/2604.02694}}

License

Apache-2.0.

DocShield-4B

README

Training data

Capabilities

Model details

Quick start

Minimal example

Inference notes

Citation

License

Explore FriendliAI today

README

Training data

Capabilities

Model details

Quick start

Minimal example

Inference notes

Citation

License