---
license: cc-by-nc-sa-4.0
# task_categories:
# - images
# - nlp
language:
- en
pretty_name: AIxpert Benchmark
size_categories:
- 10<n<10K
configs:
- config_name: image_healthcare_bias_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_bias/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_bias_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_bias/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_toxicity_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_toxicity/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_toxicity_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_toxicity/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_security_risks_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_security_risks/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_security_risks_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_security_risks/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_representation_gaps_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_representation_gaps/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_healthcare_representation_gaps_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/healthcare_representation_gaps/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_bias_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_bias/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_bias_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_bias/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_toxicity_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_toxicity/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_toxicity_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_toxicity/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_security_risks_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_security_risks/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_security_risks_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_security_risks/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_representation_gaps_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_representation_gaps/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_legal_representation_gaps_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/legal_representation_gaps/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_bias_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_bias/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_bias_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_bias/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_toxicity_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_toxicity/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_toxicity_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_toxicity/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_security_risks_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_security_risks/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_security_risks_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_security_risks/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_representation_gaps_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_representation_gaps/data_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: image_hiring_representation_gaps_csr_vqa
  data_files:
  - split: test
    path: data/image_modality/hiring_representation_gaps/data_csr_vqa.parquet
  features:
  - name: ID
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: multiple_choice_options
    dtype: string
  - name: correct_answer
    dtype: string
  - name: open_ended
    dtype: string
  - name: rationale
    dtype: string
  - name: grounding_evidence
    dtype: string
  - name: distractor_justifications
    dtype: string
  - name: csr_type
    dtype: string
  - name: image_prompt
    dtype: string
  - name: image_path
    dtype: string
  - name: metadata
    dtype: string
- config_name: nlp_hiring_bias_discrimination
  data_files:
  - split: test
    path: data/nlp_modality/answers_hiring_bias_discrimination.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: candidates
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: D
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: criteria
    dtype: string
- config_name: nlp_hiring_toxicity
  data_files:
  - split: test
    path: data/nlp_modality/answers_hiring_toxicity.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: D
    dtype: string
  - name: E
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: phrases
    dtype: string
- config_name: nlp_hiring_representation_gaps
  data_files:
  - split: test
    path: data/nlp_modality/answers_hiring_representation_gaps.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: phrases
    dtype: string
  - name: underrepresented_groups
    dtype: string
- config_name: nlp_answers_hiring_security_risks
  data_files:
  - split: test
    path: data/nlp_modality/answers_hiring_security_risks.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: risk_type
    dtype: string
  - name: risk_indicators
    dtype: string
- config_name: nlp_legal_bias_discrimination
  data_files:
  - split: test
    path: data/nlp_modality/answers_legal_bias_discrimination.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: candidates
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: D
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: criteria
    dtype: string
- config_name: nlp_legal_toxicity
  data_files:
  - split: test
    path: data/nlp_modality/answers_legal_toxicity.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: D
    dtype: string
  - name: E
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: phrases
    dtype: string
- config_name: nlp_legal_representation_gaps
  data_files:
  - split: test
    path: data/nlp_modality/answers_legal_representation_gaps.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: phrases
    dtype: string
  - name: underrepresented_groups
    dtype: string
- config_name: nlp_answers_legal_security_risks
  data_files:
  - split: test
    path: data/nlp_modality/answers_legal_security_risks.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: risk_type
    dtype: string
  - name: risk_indicators
    dtype: string
- config_name: nlp_healthcare_bias_discrimination
  data_files:
  - split: test
    path: data/nlp_modality/answers_healthcare_bias_discrimination.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: candidates
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: D
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: criteria
    dtype: string
- config_name: nlp_healthcare_toxicity
  data_files:
  - split: test
    path: data/nlp_modality/answers_healthcare_toxicity.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: D
    dtype: string
  - name: E
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: phrases
    dtype: string
- config_name: nlp_healthcare_representation_gaps
  data_files:
  - split: test
    path: data/nlp_modality/answers_healthcare_representation_gaps.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: phrases
    dtype: string
  - name: underrepresented_groups
    dtype: string
- config_name: nlp_answers_healthcare_security_risks
  data_files:
  - split: test
    path: data/nlp_modality/answers_healthcare_security_risks.parquet
  features:
  - name: id
    dtype: string
  - name: domain
    dtype: string
  - name: risk
    dtype: string
  - name: prompt
    dtype: string
  - name: text
    dtype: string
  - name: MCQ
    dtype: string
  - name: A
    dtype: string
  - name: B
    dtype: string
  - name: C
    dtype: string
  - name: answer
    dtype: string
  - name: explanation
    dtype: string
  - name: risk_type
    dtype: string
  - name: risk_indicators
    dtype: string
---
# AIXPERT

---

## License

This dataset is licensed under the **CC BY-NC-SA 4.0 License**.

## 📚 Citation

```bibtex
@article{TODO
}