| --- |
| license: apache-2.0 |
| library_name: peft |
| base_model: Qwen/Qwen3-VL-8B-Instruct |
| base_model_relation: adapter |
| datasets: |
| - ruotian/ConClick-Data |
| pipeline_tag: image-text-to-text |
| tags: |
| - gui-grounding |
| - computer-use |
| - qwen3-vl |
| - lora |
| - conclick |
| --- |
| |
| # ConClick-8B |
|
|
| Model for **ConClick: Element-wise Contrastive Learning Improves GUI Grounding**. ConClick trains on target–distractor examples with coordinate supervision, listwise ranking, and pairwise margin losses. |
| | Inference | ScreenSpot-Pro | MMBench-GUI L2 | OSWorld-G | |
| |---|---:|---:|---:| |
| | Direct | 66.03 | 86.28 | 70.20 | |
| | + ReClick | 72.36 | 87.28 | 71.76 | |
|
|
| Contains the LoRA adapter, selection head, and visual-merger state for **Qwen/Qwen3-VL-8B-Instruct**. Use the [ConClick repository](https://github.com/zhangruotian/ConClick) to install dependencies, run inference, evaluate, or train. |
