README.md

metadata

license: apache-2.0
task_categories:
  - question-answering
tags:
  - finance
  - insurance
size_categories:
  - 10K<n<100K
language:
  - en

INS-MMBench Dataset

INS-MMBench is the first comprehensive LVLMs benchmark for the insurance domain, proposed in our paper:

INS-MMBench: A Comprehensive Benchmark for Evaluating LVLMs' Performance in Insurance

📄Paper Link | 🐙GitHub Repository

---

Overview

INS-MMBench is the first comprehensive LVLMs benchmark for the insurance domain, it covers four representative insurance types: auto, property, health, and agricultural insurance and key insurance stages such as risk underwriting, risk monitoring and claim processing. INS-MMBench consists of three layers task:

• Fundamental task, which focuses on the understanding of individual insurance-related visual elements;
• Meta-task, which involves the compositional understanding of multiple insurance-related visual elements;
• Scenario task, which pertains to real-world insurance tasks requiring multi-step reasoning and decision-making.

INS-MMBench includes a total of 12,052 images, 10,372 thoroughly designed questions (including multiple-choice visual questions and free-text visual questions), comprehensively covering 5 scenario tasks, 12 meta-tasks and 22 fundamental tasks.

This dataset includes six TSV files, each corresponding to a specific task described in our paper:

File Name	Task Type
INS_MMBench_fundamental.tsv	Fundamental Task
multi_step_claim.tsv	Scenario Task - Auto Insurance Claim Processing
multi_step_liability.tsv	Scenario Task - Auto Insurance Accident Liability Determination
multi_step_health.tsv	Scenario Task - Health Insurance Risk Assessment
multi_step_property.tsv	Scenario Task - Property Insurance Risk Management
multi_step_agri.tsv	Scenario Task - Agricultural Insurance Claim Processing

---