Research map / Video / Video-language understanding
MIST: Multi-Modal Iterative Spatial-Temporal Transformer for Long-Form Video Question Answering
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use MIST instead of Temp[ATP] | CLIP-based VideoQA · AGQA v2 test | accuracy (All): 49.79 -> 54.39 | Helped |
| Use MIST instead of VGT | VideoQA · NExT-QA val | accuracy (All): 55.02 -> 57.18 | Helped |
| Use MIST instead of Temp[ATP] | CLIP-based VideoQA · STAR val | accuracy (Mean): 48.37 -> 51.13 | Helped |
| Use MIST instead of TSEA | VideoQA · Env-QA test | QA accuracy (All): 47.06 -> 48.97 | Helped |
| Use MIST instead of Temp[ATP] | CLIP-based VideoQA · AGQA v1 test | accuracy (All): 54.17 -> 58.26 | Helped |
| Add Segment Selection (ablation) | MIST · AGQA v2 | accuracy (All): 51.40 -> 54.39 | Helped |
| Add Segment Selection (ablation) | MIST · NExT-QA | accuracy (All): 53.49 -> 57.18 | Helped |
| Add Region Selection (ablation) | MIST · AGQA v2 | accuracy (All): 54.13 -> 54.39 | No clear effect |
| Add Region Selection (ablation) | MIST · NExT-QA | accuracy (All): 56.81 -> 57.18 | No clear effect |
| Use spatial-temporal self-attention instead of mean pooling (ablation) | MIST · AGQA v2 | accuracy (All): 43.79 -> 54.39 | Helped |
| Use spatial-temporal self-attention instead of mean pooling (ablation) | MIST · NExT-QA | accuracy (All): 49.12 -> 57.18 | Helped |
| Use Gumbel-Softmax selection with replacement instead of without replacement (ablation) | MIST · AGQA v2 test | accuracy (All): 54.39 -> 53.84 | Hurt |
Closest papers by meaning
- Modularized Self-Reflected Video Reasoner for Multimodal LLM with Application to Video Question Answering (ICML 2025 · similarity 0.90)
- Tool-Augmented Spatiotemporal Reasoning for Streamlining Video Question Answering Task (NeurIPS 2025 · similarity 0.89)
- MoReVQA: Exploring Modular Reasoning Models for Video Question Answering (CVPR 2024 · similarity 0.89)
- Language-aware Visual Semantic Distillation for Video Question Answering (CVPR 2024 · similarity 0.89)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.89)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.