Mapping the Concept Landscape: Structural Perception of Global Distributions for Transparent Data Pruning
2026-08-24 • Machine Learning
Machine LearningComputer Vision and Pattern Recognition
AI summaryⓘ
The authors noticed that current ways to pick important data often use simple summaries of images and captions, which can miss rare but meaningful details. They created a new method called Mapping the Concept Landscape (MCL) that turns each image and caption into detailed graphs showing things like objects, actions, and qualities. By combining these graphs, their method understands which ideas are rare across the whole dataset and picks data that covers these rare ideas well. Their approach works better than previous methods and makes it clear why certain data points were chosen.
data pruningfeature embeddingssemantic conceptsgraph representationsample selectiongreedy algorithmconcept coverageimage-caption datasetsinterpretability
Authors
Dongyue Wu, Tao Ma
Abstract
Existing data pruning methods predominantly rely on high-dimensional feature embeddings to measure sample importance. However, these compressed vectors often obscure fine-grained semantic interactions, leading to suboptimal coverage of rare semantic concepts in the pruned subsets. In this paper, we propose Mapping the Concept Landscape (MCL), a novel structural perception framework for transparent data pruning. Instead of abstract embeddings, we represent each image-caption pair as an explicit sample-level graph comprising entities, events, and attributes. By integrating these individual graphs into a comprehensive dataset-level graph, we characterize the global distribution of semantic concepts and quantify their rarity across the entire corpus. Based on this structured perception, we develop a greedy concept-coverage maximization algorithm that iteratively selects samples to maximize the marginal gain of high-value, under-represented concepts. Experimental results on various benchmarks demonstrate that our method not only achieves superior pruning efficiency compared to state-of-the-art methods but also provides a transparent and interpretable audit trail for the selection process.