README.md · submission1331/ProactiveBench at main

metadata

configs:
  - config_name: ChangeIt
    data_files:
      - split: test
        path: ChangeIt_preprocessed.jsonl
  - config_name: OcclusionDataSet-MM20
    data_files:
      - split: test
        path: OcclusionDataSet-MM20_preprocessed.jsonl
  - config_name: coco2014
    data_files:
      - split: test
        path: coco2014_preprocessed.jsonl
  - config_name: ImageNet-C
    data_files:
      - split: test
        path: ImageNet-C_preprocessed.jsonl
  - config_name: MVP-N
    data_files:
      - split: test
        path: MVP-N_preprocessed.jsonl
  - config_name: QuickDraw
    data_files:
      - split: test
        path: QuickDraw_preprocessed.jsonl
  - config_name: Realistic-Occlusion-Dataset
    data_files:
      - split: test
        path: Realistic-Occlusion-Dataset_preprocessed.jsonl
license: cc-by-nc-4.0
task_categories:
  - question-answering
language:
  - en
size_categories:
  - 10K<n<100K
tags:
  - Proactive
  - Reasoning
  - Multimodal
  - LLM
pretty_name: ProactiveBench

This is the benchmark associated with submission 1331 at NeurIPS 2025 - Dataset and Benchmarks track. The benchmark is intended to be used with the proposed submission environments (see the source code). The .jsonl files do not contain proper image paths but rather image path templates, as each .jsonl entry is a sample, and each sample corresponds to a different environment with its own images. See the submitted code README for information about dataset downloading and preprocessing, and to re-run the evaluations.