Research map / Vision-language and multimodal / Human and embodied modeling
Visual Concept-Metaconcept Learning
Chi Han, Jiayuan Mao, Chuang Gan, Josh Tenenbaum, Jiajun Wu
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 |
|---|---|---|---|
| Jointly learn visual concepts and metaconcepts with VCML | CLEVR synonym generalization · CLEVR | accuracy: $100.0_{\pm0.0}$ vs. $66.3_{\pm1.4}$ | Improved |
| Jointly learn visual concepts and metaconcepts with VCML | CLEVR same-kind generalization · CLEVR | accuracy: $99.3_{\pm1.0}$ vs. $92.3_{\pm4.9}$ | Improved |
| Jointly learn visual concepts and metaconcepts with VCML | GQA synonym generalization · GQA | accuracy: $91.1_{\pm1.7}$ vs. $83.1_{\pm1.5}$ | Improved |
| Jointly learn visual concepts and metaconcepts with VCML | GQA same-kind generalization · GQA | accuracy: $69.1_{\pm1.7}$ vs. $68.2_{\pm4.0}$ | Improved |
| Jointly learn visual concepts and metaconcepts with VCML | CUB hypernym generalization · CUB | accuracy: $94.8_{\pm1.3}$ vs. $80.1_{\pm7.3}$ | Improved |
| Use VCML instead of NS-CL | CUB meronym generalization · CUB | accuracy: $97.7_{\pm1.1}$ vs. $92.5_{\pm1.0}$ | Hurt |
| Include metaconcept questions during training (ablation) | VCML few-shot visual concept learning · CUB held-out concepts | test accuracy: 78.5\% vs. 78.5\% | Improved |
| Include metaconcept questions during training (ablation) | NS-CL few-shot visual concept learning · CUB held-out concepts | test accuracy: 79.6\% | Improved |
| Include metaconcept questions during training (ablation) | MAC few-shot visual concept learning · CUB held-out concepts | test accuracy: 60.4\% | Improved |
Closest papers by meaning
- ColorConceptBench: A Benchmark for Probabilistic Color-Concept Understanding in Text-to-Image Models (NeurIPS 2026 · similarity 0.87)
- Language-Informed Visual Concept Learning (ICLR 2024 · similarity 0.87)
- Bridging the gap to real-world language-grounded visual concept learning (NeurIPS 2025 · similarity 0.86)
- Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.85)
- Visual Chain-of-Thought Prompting for Knowledge-Based Visual Reasoning (AAAI 2024 · similarity 0.84)
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · 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.
