Datasets:

Modalities:
Image
Sub-tasks:
fact-checking
Languages:
English
ArXiv:
Libraries:
Datasets
License:

Update task categories to include image-text-to-text

#2
by nielsr HF Staff - opened
Files changed (1) hide show
  1. README.md +8 -7
README.md CHANGED
@@ -5,16 +5,18 @@ language:
5
  - en
6
  license: cc-by-4.0
7
  multilinguality: monolingual
8
- pretty_name: SciVer
9
  size_categories:
10
  - 1K<n<10K
11
  source_datasets:
12
  - original
13
  task_categories:
14
  - text-classification
 
15
  task_ids:
16
  - fact-checking
 
17
  ---
 
18
  # SCIVER: A Benchmark for Multimodal Scientific Claim Verification
19
 
20
  <p align="center">
@@ -23,7 +25,6 @@ task_ids:
23
  <a href="https://huggingface.co/datasets/chengyewang/SciVer">πŸ€— Data</a>
24
  </p>
25
 
26
-
27
  ## πŸ“° News
28
  - [May 15, 2025] SciVer has been accepted by ACL 2025 Main!
29
 
@@ -47,7 +48,7 @@ task_ids:
47
  - Analytical
48
  - πŸ“š Context includes **text paragraphs, multiple tables, and charts**
49
  - πŸ” Labels: `Entailed`, `Refuted`
50
- - πŸ“ˆ Evaluated across **21 leading foundation models**, including GPT-4o, Gemini, Claude 3.5, Qwen2.5-VL, LLaMA-3.2-Vision, etc.
51
  - βš–οΈ Includes **step-by-step rationale** and **automated accuracy evaluation**
52
 
53
  ------
@@ -65,8 +66,9 @@ Each SCIVER sample includes:
65
 
66
  1. **Direct Reasoning** – extract simple facts
67
  2. **Parallel Reasoning** – synthesize info from multiple sources
68
- 3. **Sequential Reasoning** – perform step-by-step inference
69
- 4. **Analytical Reasoning** – apply domain expertise and logic
 
70
 
71
  ------
72
 
@@ -153,5 +155,4 @@ If you use our work and are inspired by our work, please consider cite us:
153
  primaryClass={cs.CL},
154
  url={https://arxiv.org/abs/2506.15569},
155
  }
156
- ```
157
-
 
5
  - en
6
  license: cc-by-4.0
7
  multilinguality: monolingual
 
8
  size_categories:
9
  - 1K<n<10K
10
  source_datasets:
11
  - original
12
  task_categories:
13
  - text-classification
14
+ - image-text-to-text
15
  task_ids:
16
  - fact-checking
17
+ pretty_name: SciVer
18
  ---
19
+
20
  # SCIVER: A Benchmark for Multimodal Scientific Claim Verification
21
 
22
  <p align="center">
 
25
  <a href="https://huggingface.co/datasets/chengyewang/SciVer">πŸ€— Data</a>
26
  </p>
27
 
 
28
  ## πŸ“° News
29
  - [May 15, 2025] SciVer has been accepted by ACL 2025 Main!
30
 
 
48
  - Analytical
49
  - πŸ“š Context includes **text paragraphs, multiple tables, and charts**
50
  - πŸ” Labels: `Entailed`, `Refuted`
51
+ - πŸ“ˆ Evaluated across **21 leading foundation models**, including o4-mini, GPT-4o, Gemini, Claude 3.5, Qwen2.5-VL, LLaMA-3.2-Vision, etc.
52
  - βš–οΈ Includes **step-by-step rationale** and **automated accuracy evaluation**
53
 
54
  ------
 
66
 
67
  1. **Direct Reasoning** – extract simple facts
68
  2. **Parallel Reasoning** – synthesize info from multiple sources
69
+ 3. **Sequential Reasoning** – synthesize info from multiple sources
70
+ 4. **Sequential Reasoning** – perform step-by-step inference
71
+ 5. **Analytical Reasoning** – apply domain expertise and logic
72
 
73
  ------
74
 
 
155
  primaryClass={cs.CL},
156
  url={https://arxiv.org/abs/2506.15569},
157
  }
158
+ ```