Research map / Video / Video-language understanding
FunnyNet-W: Multimodal Learning of Funny Moments in Videos in the Wild
Zhi-Song Liu, Robin Courant, Vicky S. Kalogeiton
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 VideoMAE instead of Timesformer visual encoder (gt text) (ablation) | FunnyNet-W · Friends | F1: 84.9 -> 87.2 | Improved |
| use LlaMa-2 instead of Bert text encoder (auto text) (ablation) | FunnyNet-W · Friends | F1: 85.3 -> 88.2 | Improved |
| use GPT2 instead of Bert text encoder (ablation) | FunnyNet-W · Friends | F1: 85.3 -> 85.2 | No clear effect |
| use automatically generated text instead of ground-truth text (ablation) | FunnyNet-W · Friends | F1: 89.3 -> 88.2 | Hurt |
| use BYOL-A instead of BEATs audio encoder (ablation) | FunnyNet-W · Friends | F1: 78.2 -> 88.2 | Improved |
| add text to visual+audio (ablation) | FunnyNet-W · Friends | F1: 84.3 -> 88.2 | Improved |
| add audio to visual+text (ablation) | FunnyNet-W · Friends | F1: 74.9 -> 88.2 | Improved |
| add visual to audio+text (ablation) | FunnyNet-W · Friends | F1: 84.5 -> 88.2 | Improved |
| add self and cross attention in CAF (ablation) | FunnyNet-W · Friends | F1: 82.4 -> 88.2 | Improved |
| use CAF instead of MMCA fusion (ablation) | FunnyNet-W · Friends | F1: 87.0 -> 88.2 | Improved |
| add self-supervised audiovisual loss L_ss (ablation) | FunnyNet-W · Friends | F1: 70.9 -> 88.2 | Improved |
| add Koleo loss (ablation) | FunnyNet-W · Friends | F1: 88.2 -> 86.7 | Hurt |
Closest papers by meaning
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.84)
- MMOU: A Massive Multi-Task Omni Understanding and Reasoning Benchmark for Long and Complex Real-World Videos (NeurIPS 2026 · similarity 0.84)
- SAMA: Towards Multi-Turn Referential Grounded Video Chat with Large Language Models (NeurIPS 2025 · similarity 0.84)
- HourVideo: 1-Hour Video-Language Understanding (NeurIPS 2024 · similarity 0.84)
- Watch and Listen: Understanding Audio-Visual-Speech Moments with Multimodal LLM (NeurIPS 2025 · similarity 0.84)
- FAVOR-Bench: A Comprehensive Benchmark for Fine-Grained Video Motion Understanding (NeurIPS 2025 · similarity 0.84)
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.
