MMDU: AMulti-TurnMulti-ImageDialog UnderstandingBenchmarkand Instruction-Tuning DatasetforLVLMs

Feb-8-2026, 01:15:40 GMT–Neural Information Processing Systems

Existing LVLM benchmarks primarily focus onsingle-choice questions orshort-form responses, which donotadequately assess the capabilities ofLVLMs inreal-world human-AI interaction applications.

large language model, machine learning, natural language, (21 more...)

Neural Information Processing Systems

Feb-8-2026, 01:15:40 GMT

Conferences PDF

Add feedback

Country:
- Asia > China > Jiangsu Province > Changzhou (0.04)

Genre:
- Research Report (0.46)

Technology:
- Information Technology > Artificial Intelligence
  - Machine Learning > Neural Networks
    - Deep Learning (0.52)
  - Natural Language > Large Language Model (0.74)

Duplicate Docs Excel Report

Title
MMDU: A Multi-T urn Multi-Image Dialog Understanding Benchmark and Instruction-T uning Dataset for L VLMs

Similar Docs Excel Report more

Title	Similarity	Source
None found