Research map / Vision-language and multimodal / Multimodal LLMs
PACS: A Dataset for Physical Audiovisual Commonsense Reasoning
Samuel Yu, Peter Wu, Paul Pu Liang, Ruslan Salakhutdinov, Louis-Philippe Morency
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 |
|---|---|---|---|
| add audio input for human annotators (ablation) | Human · PACS test | accuracy (%): 90.5 -> 96.3 | Improved |
| add audio input (ablation) | Merlot Reserve (Large) · PACS test | accuracy (%): 68.4 -> 70.1 | Improved |
| add audio input (ablation) | Merlot Reserve (Base) · PACS test | accuracy (%): 64.0 -> 66.5 | Improved |
| add audio input (ablation) | CLIP/AudioCLIP · PACS test | accuracy (%): 56.3 -> 60.0 | Improved |
| add audio input (ablation) | Late Fusion · PACS test | accuracy (%): 52.5 -> 55.0 | Improved |
| use only audio plus question instead of all modalities (ablation) | late-fusion bias model · PACS test | accuracy (%): 50.9 -> 55.0 | No clear effect |
| scale Merlot Reserve from Base to Large | Merlot Reserve · PACS test (with audio) | accuracy (%): 66.5 -> 70.1 | Improved |
| change task from physical commonsense QA to material classification | AudioCLIP · PACS test | accuracy (%): 60.0 -> 75.9 | Improved |
| change task from physical commonsense QA to material classification | Late Fusion · PACS test | accuracy (%): 55.0 -> 67.4 | Improved |
Closest papers by meaning
- Counterfactual Debiasing for Physical Audiovisual Commonsense Reasoning (AAAI 2025 · similarity 0.91)
- Mellow: a small audio language model for reasoning (NeurIPS 2025 · similarity 0.86)
- Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation (ICML 2025 · similarity 0.85)
- STAR-Bench: Probing Deep Spatio-Temporal Reasoning as Audio 4D Intelligence (ICLR 2026 · similarity 0.85)
- SAVVY: Spatial Awareness via Audio-Visual LLMs through Seeing and Hearing (NeurIPS 2025 · similarity 0.85)
- SciVideoBench: Benchmarking Scientific Video Reasoning in Large Multimodal Models (ICML 2026 · 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.
