Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.08175 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910202456965120 |
|---|---|
| author | Ghosh, Archishman Roy, Abhinaba Herremans, Dorien |
| author_facet | Ghosh, Archishman Roy, Abhinaba Herremans, Dorien |
| contents | While significant progress has been made in Video Question Answering and cross-modal understanding, causal reasoning about how visual dynamics drive musical structure in music videos remains under-explored. We introduce KARMA-MV, a large-scale multiple-choice QA dataset derived from 2,682 YouTube music videos, designed to test models' ability to integrate temporal audio-visual cues and reason about visual-to-musical influence across reasoning, prediction, and counterfactual questions. Unlike traditional datasets requiring manual annotation, KARMA-MV leverages LLM reasoning for scalable generation and validation, yielding 37,737 MCQs. We propose a causal knowledge graph (CKG) approach that augments vision-language models (VLMs) with structured retrieval of cross-modal dependencies. Experiments on state-of-the-art VLMs and LLMs show consistent gains from CKG grounding -- especially for smaller models -- establishing the value of explicit causal structure for music-video reasoning. KARMA-MV provides a new benchmark for advancing causal audio-visual understanding beyond correlation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_08175 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | KARMA-MV: A Benchmark for Causal Question Answering on Music Videos Ghosh, Archishman Roy, Abhinaba Herremans, Dorien Computer Vision and Pattern Recognition Artificial Intelligence 68T01 I.2.6; I.2.10; H.3.3 While significant progress has been made in Video Question Answering and cross-modal understanding, causal reasoning about how visual dynamics drive musical structure in music videos remains under-explored. We introduce KARMA-MV, a large-scale multiple-choice QA dataset derived from 2,682 YouTube music videos, designed to test models' ability to integrate temporal audio-visual cues and reason about visual-to-musical influence across reasoning, prediction, and counterfactual questions. Unlike traditional datasets requiring manual annotation, KARMA-MV leverages LLM reasoning for scalable generation and validation, yielding 37,737 MCQs. We propose a causal knowledge graph (CKG) approach that augments vision-language models (VLMs) with structured retrieval of cross-modal dependencies. Experiments on state-of-the-art VLMs and LLMs show consistent gains from CKG grounding -- especially for smaller models -- establishing the value of explicit causal structure for music-video reasoning. KARMA-MV provides a new benchmark for advancing causal audio-visual understanding beyond correlation. |
| title | KARMA-MV: A Benchmark for Causal Question Answering on Music Videos |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence 68T01 I.2.6; I.2.10; H.3.3 |
| url | https://arxiv.org/abs/2605.08175 |