Dissecting Generalized Category Discovery: Multiplex Consensus under Self-Deconstruction

2025-08-14

AI Summary

Generalized Category Discovery (GCD) aims to jointly recognize both known and unknown categories, yet existing methods predominantly optimize objective functions without modeling the human cognitive decomposition process. This paper proposes ConGCD, the first GCD framework grounded in cognitive principles: it employs a self-decomposition mechanism to factor images into visual primitives and reconstruct high-order semantic representations; introduces leader and contextual consensus units, coordinated by a dynamic consensus scheduler for multi-path consensus integration; and unifies primitive binding, distribution-invariant modeling, and semantic reconstruction. Evaluated on both coarse- and fine-grained benchmarks, ConGCD significantly outperforms state-of-the-art methods, demonstrating the effectiveness and interpretability of consensus-aware learning in GCD.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Human perceptual systems excel at inducing and recognizing objects across both known and novel categories, a capability far beyond current machine learning frameworks. While generalized category discovery (GCD) aims to bridge this gap, existing methods predominantly focus on optimizing objective functions. We present an orthogonal solution, inspired by the human cognitive process for novel object understanding: decomposing objects into visual primitives and establishing cross-knowledge comparisons. We propose ConGCD, which establishes primitive-oriented representations through high-level semantic reconstruction, binding intra-class shared attributes via deconstruction. Mirroring human preference diversity in visual processing, where distinct individuals leverage dominant or contextual cues, we implement dominant and contextual consensus units to capture class-discriminative patterns and inherent distributional invariants, respectively. A consensus scheduler dynamically optimizes activation pathways, with final predictions emerging through multiplex consensus integration. Extensive evaluations across coarse- and fine-grained benchmarks demonstrate ConGCD's effectiveness as a consensus-aware paradigm. Code is available at github.com/lytang63/ConGCD.

Problem

• Bridges gap between human and machine category discovery
• Decomposes objects into visual primitives for understanding
• Captures class-discriminative patterns via consensus units

Innovation

• Primitive-oriented representations via semantic reconstruction
• Dominant and contextual consensus units integration
• Dynamic consensus scheduler optimizes activation pathways