GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
densereward
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
densereward
Model Tree
Qwen/Qwen3-VL-4B-Instruct
Input Modalities
Output Modalities
Supported Functionality
<think>WORD</think> block followed by one float with three
decimals, e.g. <think>\ncorrect\n</think>\n\n0.374.The model was trained with a fixed system prompt and a strict output
format. Use the same system prompt at inference (system_prompt.txt in this
directory). It instructs:
<think>...</think>, from this
vocabulary:
correct — progress on track (reward increasing/holding)miss — gripper missed the grasp and reward just droppedcollision — an unintended collision just dropped the rewardfall — the held object fell and reward just droppedsmooth — motion stalled / non-smooth behaviour dropped the rewardfailure — an unclassified failure event just dropped the reward[0.000, 1.000] (three decimals).Canonical output (note the blank line after </think>):
text
<think>correct</think>0.374
The user turn must contain the 3 images followed by the task text, in the same format used during training (
<image><image><image>{task}).
Requires transformers>=4.57, qwen_vl_utils>=0.0.14, torch, accelerate.
Tested environment: Python 3.12, CUDA 12.8,
torch==2.8.0+cu128,torchvision==0.23.0+cu128,transformers==5.2.0,accelerate==1.13.0,qwen-vl-utils==0.0.14.bash
conda create -n densereward python=3.12 -yconda activate denserewardpip install torch==2.8.0 torchvision==0.23.0 --index-url https://download.pytorch.org/whl/cu128pip install "transformers==5.2.0" "accelerate==1.13.0" "qwen-vl-utils==0.0.14" pillow numpy
python
import reimport torchfrom transformers import AutoModelForImageTextToText, AutoProcessorfrom qwen_vl_utils import process_vision_infoMODEL_DIR = "densereward/densereward-3frame-thinking" # or a local path to this checkpoint# The exact thinking system prompt used in training ships with the model:SYSTEM_PROMPT = open(f"{MODEL_DIR}/system_prompt.txt").read().strip()model = AutoModelForImageTextToText.from_pretrained(MODEL_DIR, torch_dtype=torch.bfloat16, device_map="auto")processor = AutoProcessor.from_pretrained(MODEL_DIR)# 3 chronological frames: oldest -> currentframes = ["file:///path/to/frame_t-2.png","file:///path/to/frame_t-1.png","file:///path/to/frame_t.png",]messages = [{"role": "system", "content": SYSTEM_PROMPT},{"role": "user","content": [{"type": "image", "image": frames[0]},{"type": "image", "image": frames[1]},{"type": "image", "image": frames[2]},{"type": "text", "text": "put the black bowl on the plate"},],},]text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)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=32, do_sample=False) # greedygen = out[:, inputs.input_ids.shape[1]:]raw = processor.batch_decode(gen, skip_special_tokens=True)[0].strip()print(raw)word = re.search(r"<think>\s*(\w+)\s*</think>", raw)reward = re.search(r"</think>\s*([01](?:\.\d+)?)", raw)print("reason:", word.group(1) if word else None,"| reward:", float(reward.group(1)) if reward else None)
Notes:
max_new_tokens>=32: the output includes the <think> block and the
reward.[0, 1] and handle malformed output.The <think>WORD</think> block is part of the trained assistant output,
driven by the system prompt and thinking-labelled training data — it is not
ms-swift's native --enable_thinking mode (enable_thinking was false for this
run). At inference you only need the bundled system prompt; the model produces the
<think>...</think> block itself.
The release includes a minimal args.json ({model_type: qwen3_vl, swift_version})
so ms-swift's PtEngine / TransformersEngine can auto-detect the model type for
this local directory. Load it as the base model with no adapter; pass the
bundled system prompt and a 3-image user turn.
Compatibility note: ms-swift 4.2.x targets
transformers>=4.57,<5. Under a much newer transformers (e.g. 5.x) the swift template/prompt composition can misbehave even though weights load fine — prefer thetransformerspath above, or a swift-matched transformers version, for swift-based inference.
Apache License 2.0 (see LICENSE). The base model
Qwen/Qwen3-VL-4B-Instruct
is also Apache-2.0. This fine-tune was produced at the University of North
Carolina at Chapel Hill.
bibtex
@article{fang2026densereward,title={DenseReward: Dense Reward Learning via Failure Synthesis for Robotic Manipulation},author={Fang, Yu and Dong, Wanxi and Liu, Jiaqi and Yang, Yue and Huo, Mingxiao and Mu, Yao and Yao, Huaxiu and Li, Li Erran and Szafir, Daniel and Ding, Mingyu},journal={arXiv preprint arXiv:2607.13033},year={2026}}