Research map / Label-efficient and robust learning / Recognition and long-tail
Mapping the Concept Landscape: Structural Perception of Global Distributions for Transparent Data Pruning
Dongyue Wu, Tao Ma
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 greedy marginal-gain selection instead of static importance ranking (ablation) | MCL · COCO 2017 | AP | Improved |
| select with concept graphs instead of embedding-based diversity (PFB, DivBS) | MCL · COCO 2017 (70% data) | AP50:95: 39.69 -> 40.12 | Improved |
| prune with MCL instead of random selection | LLaVA-v1.5-7B · LLaVA-1.5-mix-665k (50k) | Rel. performance: 95.7% -> 99.0% | Improved |
| prune with MCL instead of DataTailor | LLaVA-v1.5-7B · LLaVA-1.5-mix-665k (50k) | Rel. performance: 98.6% -> 99.0% | Improved |
| raise budget from 50k to 133k with MCL | LLaVA-v1.5-7B · LLaVA-1.5-mix-665k | Rel. performance: 99.0% -> 100.5% | Improved |
| prune with MCL versus ICONS at 133k | LLaVA-v1.5-7B · LLaVA-1.5-mix-665k (133k) | Rel. performance: 99.9% -> 100.5% | Improved |
| prune COCO to 70% with MCL versus full data | DETR · COCO 2017 | AP50:95: 40.06 -> 40.12 | No clear effect |
| prune COCO to 70% with random selection | DETR · COCO 2017 | AP50:95: 40.06 -> 38.26 | Hurt |
| prune COCO to 70% with EL2N | DETR · COCO 2017 | AP50:95: 40.06 -> 34.35 | Hurt |
| MCL pruning pipeline versus TIVE and DataTailor cost | MCL · LLaVA-1.5-mix-665k | pruning time: 1.7 hours | Improved |
Closest papers by meaning
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.86)
- You Can’t Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning (NeurIPS 2026 · similarity 0.86)
- Don't Just Chase “Highlighted Tokens” in MLLMs: Revisiting Visual Holistic Context Retention (NeurIPS 2025 · similarity 0.85)
- Effective pruning of web-scale datasets based on complexity of concept clusters (ICLR 2024 · similarity 0.85)
- Sieve: Multimodal Dataset Pruning using Image Captioning Models (CVPR 2024 · similarity 0.85)
- Concept-Aware Batch Sampling Improves Language-Image Pretraining (CVPR 2026 · similarity 0.85)
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.
