Research map / Vision-language and multimodal / Human and embodied modeling
Bongard-LOGO: A New Benchmark for Human-Level Concept Learning and Reasoning
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 |
|---|---|---|---|
| replace MoCo pre-training with program-synthesis pre-training (symbolic action supervision) (ablation) | Meta-Baseline · Bongard-LOGO test sets | accuracy (%): FF 65.9 -> 68.2; NV 64.7 -> 71.5 | Helped |
| add MoCo unsupervised pre-training to Meta-Baseline-SC (ablation) | Meta-Baseline · Bongard-LOGO test sets | test accuracy (%): train 75.4 -> 81.2; FF 66.3 -> 65.9; BA 73.3 -> 72.2 | No clear effect |
| use meta-learning (ProtoNet) instead of non-meta-learning CNN-Baseline | few-shot concept learning · Bongard-LOGO FF test | accuracy (%): 51.9 -> 64.6 | Helped |
| restrict benchmark to free-form problems only (remove context-dependent and analogy properties) (ablation) | all methods · Bongard-LOGO FF test | best test accuracy (%): 66.3 -> 74.5 | Helped |
| reduce backbone model size (ablation) | ProtoNet, Meta-Baseline-SC · Bongard-LOGO test sets | test accuracy | Hurt |
| use MDN with mixture of Gaussians instead of L2 prediction for continuous action arguments (ablation) | Meta-Baseline-PS · Bongard-LOGO | accuracy | Helped |
Closest papers by meaning
- Bongard in Wonderland: Visual Puzzles that Still Make AI Go Mad? (ICML 2025 · similarity 0.92)
- Bongard-OpenWorld: Few-Shot Reasoning for Free-form Visual Concepts in the Real World (ICLR 2024 · similarity 0.91)
- Reasoning Limitations of Multimodal Large Language Models. A case study of Bongard Problems (ICML 2025 · similarity 0.89)
- Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.88)
- MARVEL: Multidimensional Abstraction and Reasoning through Visual Evaluation and Learning (NeurIPS 2024 · similarity 0.88)
- Concept-RuleNet: Grounded Multi-Agent Neurosymbolic Reasoning in Vision Language Models (AAAI 2026 · 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.