GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
jucamohedano
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
jucamohedano
Model Tree
Qwen/Qwen3-VL-4B-Instruct
Input Modalities
Output Modalities
Supported Functionality
Qwen/Qwen3-VL-4B-Instructtask_type: CAUSAL_LM)0.05, exact/fuzzy 0.70,
specificity-weighted hF 0.15, path-match 0.05, aggregation 0.05); a missing \boxed{}
answer returns 0.00.| metric | start | end | notes |
|---|---|---|---|
| policy entropy | 0.431 | 0.253 | min 0.242 |
| KL to reference | 0.161 | 1.875 | max 1.875 |
| training reward | 0.197 | 0.056 | — |
| response length | 230.977 | 232.820 | tokens |
Training reward moved 0.197 → 0.056 while held-out exact-match went 0.141 → 0.141 (essentially flat). Consistent with the thesis: outcome-only GRPO does not raise validation accuracy above the prompt-only elicitation floor. As a small-KL LoRA update the policy stays close to the reference model.
python
from peft import PeftModelfrom transformers import AutoModelForImageTextToText, AutoProcessorbase = AutoModelForImageTextToText.from_pretrained("Qwen/Qwen3-VL-4B-Instruct", trust_remote_code=True, device_map="auto")model = PeftModel.from_pretrained(base, "jucamohedano/qwen3-vl-4b-oven-grpo-traversal-structured-exact")processor = AutoProcessor.from_pretrained("Qwen/Qwen3-VL-4B-Instruct", trust_remote_code=True)
Research artifact for reproducing the thesis's (largely negative) reinforcement-learning results on OVEN and for studying GRPO training dynamics on a multimodal open-world task. Not tuned or recommended for production classification.
MSc thesis, University of Trento (Juan Camacho Mohedano). Training: verl (branch grpo-oven-v080);
reward in verl/utils/reward_score/oven_boxed.py; data by oven-mllm-eval/scripts/build_verl_oven_parquet.py.
wandb run: offline-run-20260704_140349-qwen3-vl-4b-oven-grpo-traversal-structured-exact-clean-seed42.
| 0.141 |
| 0.141 |
| held-out |