Datasets:

Modalities:
Image
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
marc-thibault-h commited on
Commit
01746ac
·
verified ·
1 Parent(s): 33ba62b

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -5
README.md CHANGED
@@ -29,7 +29,6 @@ configs:
29
  ---
30
 
31
 
32
-
33
  # WebClick: A Multimodal Localization Benchmark for Web-Navigation Models
34
  We introduce WebClick, a high-quality benchmark dataset for evaluating navigation and localization capabilities of multimodal models and agents in Web environments. WebClick features 1,639 English-language web screenshots from over 100 websites paired with precisely annotated natural-language instructions and pixel-level click targets, in the same format as the widely-used screenspot benchmark.
35
 
@@ -115,13 +114,14 @@ From the table we can observe that the models mostly underperform on WebClick co
115
 
116
  | **Model** | **WebClick (ours)** | Screenspot | Screenspot V2 |
117
  |-------------------------------|----------------------------|------------|---------------|
118
- | osunlp/UGround-V1-2B [3] | 71.69% | 71.12% | 79.31% |
119
  | osunlp/UGround-V1-7B [3] | 82.37% | 85.69% | 84.26% |
120
  | Qwen/Qwen2.5-VL-3B-Instruct [4] | 71.15% | 82.78% | 84.34% |
121
- | Qwen/Qwen2.5-VL-7B-Instruct [4] | 75.28% | 85.84% | 88.04% |
122
  | ByteDance-Seed/UI-TARS-2B-SFT [5] | 64.23% | 66.82% | 69.39% |
123
- | ByteDance-Seed/UI-TARS-7B-DPO [5] | 80.67% | 83.56% | 86.55% |
124
- | H-1 | 84.56% | 85.53% | 87.25% |
 
125
  | | | | |
126
 
127
 
 
29
  ---
30
 
31
 
 
32
  # WebClick: A Multimodal Localization Benchmark for Web-Navigation Models
33
  We introduce WebClick, a high-quality benchmark dataset for evaluating navigation and localization capabilities of multimodal models and agents in Web environments. WebClick features 1,639 English-language web screenshots from over 100 websites paired with precisely annotated natural-language instructions and pixel-level click targets, in the same format as the widely-used screenspot benchmark.
34
 
 
114
 
115
  | **Model** | **WebClick (ours)** | Screenspot | Screenspot V2 |
116
  |-------------------------------|----------------------------|------------|---------------|
117
+ | osunlp/UGround-V1-2B [3] | 71.69% | 77.12% | 79.31% |
118
  | osunlp/UGround-V1-7B [3] | 82.37% | 85.69% | 84.26% |
119
  | Qwen/Qwen2.5-VL-3B-Instruct [4] | 71.15% | 82.78% | 84.34% |
120
+ | Qwen/Qwen2.5-VL-7B-Instruct [4] | 74.37% | 85.53% | 88.04% |
121
  | ByteDance-Seed/UI-TARS-2B-SFT [5] | 64.23% | 66.82% | 69.39% |
122
+ | ByteDance-Seed/UI-TARS-7B-DPO [5] | 80.67% | 84.20% | 86.70% |
123
+ | Holo1-3B | 81.50% | 86.01% | 87.33% |
124
+ | Holo1-7B | 84.03% | 87.42% | 89.85% |
125
  | | | | |
126
 
127