Downloads · 30 days
0
qwen2vl-7b-vlreward-lora
Hanzi7na/qwen2vl-7b-vlreward-lora
qwen2vl-7b-vlreward-lora is a machine learning model from Hanzi7na. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as other.
This repository contains the trained LoRA + scalar score head checkpoint for a VL reward model based on Qwen2-VL-7B-Instruct.
Access
Public
Updated May 19, 2026
Repo size
23.2 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt23.2 MB · 100%
At a glance
- License
- other
- Access
- Public
- Created
- May 19, 2026
- Updated
- May 19, 2026
- SHA
- 3abcc721
Base models
Model card
From the Hugging Face model README
Qwen2-VL-7B VL Reward Model LoRA
This repository contains the trained LoRA + scalar score head checkpoint for a VL reward model based on Qwen2-VL-7B-Instruct.
Result
- Base model: Qwen2-VL-7B-Instruct
- Training data: H4 RLAIF-V + RLHF-V mixed preference pairs
- Evaluation benchmark: VLRewardBench
- Score_base: 43.14%
- Score_reward: 66.56%
Files
reward_lora_best.pt: best LoRA + score head checkpointqwen2vl_reward.py: training/evaluation script used for this checkpointtrain_history.json: training history if available
VLRewardBench was used only for evaluation, not for training.
- License
- other
- Created
- May 19, 2026
- Updated
- May 19, 2026