GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Gyubeum
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
Gyubeum
Model Tree
Gyubeum/Qwen3-VL-8B-Instruct-UI-Genie
Input Modalities
Output Modalities
Supported Functionality
Drop-in scalar reward signal for GUI agent training (e.g. PPO/GRPO, RLHF) or best-of-N selection. Unlike the discrete <|+|> / <|-|> classifier, this model outputs a continuous unbounded reward that is directly usable as a value signal.
bash
pip install transformers torch pillow safetensors
Note: Use HuggingFace
transformersdirectly — vLLM'sQwen3VLForConditionalGenerationloader does not support the additionalscore.weighttensor.
python
import jsonimport osimport torchfrom safetensors import safe_openfrom transformers import AutoProcessor, Qwen3VLForConditionalGenerationfrom PIL import ImageMODEL_PATH = "Gyubeum/Qwen3-VL-8B-Instruct-UI-Genie-scoring"MAX_LEN = 8192DEVICE = "cuda"# HF tolerates the extra `score.weight` tensor in the checkpoint (warning,# harmless) — it just isn't wired into the base CausalLM class, so we load# it ourselves into a separate linear head below.model = Qwen3VLForConditionalGeneration.from_pretrained(MODEL_PATH,torch_dtype=torch.bfloat16,low_cpu_mem_usage=True,).to(DEVICE).eval()processor = AutoProcessor.from_pretrained(MODEL_PATH, max_pixels=1_048_576)tokenizer = processor.tokenizerdef _load_score_head(model_path: str) -> torch.nn.Linear:from huggingface_hub import hf_hub_downloadindex_path = hf_hub_download(model_path, "model.safetensors.index.json")with open(index_path) as f:weight_map = json.load(f)["weight_map"]shard_path = hf_hub_download(model_path, weight_map["score.weight"])with safe_open(shard_path, framework="pt") as f:weight = f.get_tensor("score.weight")head = torch.nn.Linear(weight.shape[1], weight.shape[0], bias=False)with torch.no_grad():head.weight.copy_(weight)return head.to(DEVICE, dtype=torch.bfloat16)score_head = _load_score_head(MODEL_PATH)def score_action(prefix_messages, response_text, images=None):"""Score a GUI agent action.Args:prefix_messages: Chat messages up to (not including) the assistant turn.E.g. [{"role": "system", "content": "..."},{"role": "user", "content": [{"type":"text","text":"..."},{"type":"image"}]}]response_text: The assistant tool-call response to score.images: List of PIL.Image objects matching image placeholders.Returns:float: Scalar BT reward (higher = better action)."""structured_msgs = list(prefix_messages) + [{"role": "assistant", "content": response_text.strip()}]# Mirrors training: chat template closes the assistant turn with# <|im_end|>; the score head reads the hidden state at that position.full_text = processor.apply_chat_template(structured_msgs, tokenize=False, add_generation_prompt=False)inputs = processor(text=[full_text],images=images or None,truncation=True,max_length=MAX_LEN,return_tensors="pt",).to(DEVICE)with torch.inference_mode():outputs = model(**inputs, output_hidden_states=True, return_dict=True)last_hidden = outputs.hidden_states[-1] # (1, seq_len, hidden_dim)last_idx = inputs["attention_mask"].sum(dim=1) - 1pooled = last_hidden[0, last_idx[0], :] # (hidden_dim,)reward = score_head(pooled.unsqueeze(0).to(score_head.weight.dtype))return reward.squeeze().float().item()# Example: compare two candidate actionsscreenshot = Image.open("screenshot.png").convert("RGB")SYSTEM_PROMPT = "You are a helpful assistant." # use full mobile_use tool spec in practicemessages = [{"role": "system", "content": SYSTEM_PROMPT},{"role": "user", "content": [{"type": "text", "text": "The user query: tap the search button\n"},{"type": "image"},]},]action_a = '{"name": "mobile_use", "arguments": {"action": "click", "coordinate": [540, 120]}}'action_b = '{"name": "mobile_use", "arguments": {"action": "click", "coordinate": [100, 800]}}'score_a = score_action(messages, action_a, images=[screenshot])score_b = score_action(messages, action_b, images=[screenshot])print(f"Action A: {score_a:.4f}")print(f"Action B: {score_b:.4f}")print(f"Preferred: {'A' if score_a > score_b else 'B'}")
rm_evalbash
python eval_rm.py \--rm_path Gyubeum/Qwen3-VL-8B-Instruct-UI-Genie-scoring \--datasets ui-genie \--mode bt \--uigenie_jsonl /path/to/reward_data_rm_pairs_last5.jsonl \--uigenie_images_dir /path/to/images \--output_dir results/
Pairwise accuracy on held-out (chosen, rejected) UI-Genie pairs:
| Dataset | Pairs | Pairwise accuracy | Source |
|---|---|---|---|
| UI-Genie (held-out) | 1000 | 90.6% (906/1000, 3 ties) | Fresh rm_eval run, 2026-08-11, on this uploaded checkpoint |
| UI-Genie (held-out) | 1000 | 90.5% (905/1000, 3 ties) | rm_eval run, 2026-05-01, on this exact checkpoint (mtime-verified) |
| AndroidFlux (OOD multi-agent replay) | 203 | 61.1% (124/203) | rm_eval run, 2026-04-30, on this exact checkpoint (mtime-verified) |
The two UI-Genie runs agree to within 1 pair (906 vs. 905 correct, identical tie count), confirming the fix reproduces consistently.
AndroidFlux is out-of-domain relative to the UI-Genie training data (different agents, different action distribution) and is reported for reference, not as the primary validation signal.
| Field | Value |
|---|---|
| Base model | Gyubeum/Qwen3-VL-8B-Instruct-UI-Genie |
| Training method | Bradley-Terry pairwise loss (LoRA, merged) |
| Architecture | Qwen3VLForConditionalGeneration + score.weight linear head |
| Task type | seq_cls (regression, num_labels=1) |
| Score | Last non-padding token hidden state → linear head → scalar |
| dtype | bfloat16 |
<|+|> / <|-|>) this model's score.weight was fine-tuned on top of.bibtex
@misc{qwen3technicalreport,title={Qwen3 Technical Report},author={Qwen Team},year={2025},eprint={2505.09388},archivePrefix={arXiv},primaryClass={cs.CL},}