CoGe-GCD Model for Category Discovery
CoGe-GCD: Reframing Generalized Category Discovery with Compositional Generalization
CoGe-GCD can be integrated into existing GCD frameworks without modifying heads or losses, offering practical improvements for researchers working with category discovery.
CoGe-GCD improves Generalized Category Discovery by using compositional generalization with two stages.
Compositional Perception structures patch tokens by mapping them to primitives and refining embeddings.Generalizing Induction applies structure-preserving calibration over spatial relations.
On standard benchmarks, it improves all-class accuracy, unknown-class number estimation, and geometric quality with minimal computational overhead.
CoGe-GCD: Reframing Generalized Category Discovery with Compositional Generalization
Generalized Category Discovery (GCD) assigns unlabeled instances, mixed with labeled data, to known or novel categories, requiring human-like compositional reasoning: reusing primitives learned from known classes and deciding when new combinations imply new categories. Existing GCD methods operate on unstructured token features and struggle to extrapolate to novel compositions. We propose CoGe-GCD, which rethinks GCD through compositional generalization with two coupled stages. (i) Compositional Perception structures patch tokens by mapping them to a small vocabulary of primitives and refining token embeddings via competitive token-primitive assignment and information passing, yielding coherent groups for discovery. (ii) Generalizing Induction exploits the induced geometric structure and applies a structure-preserving calibration over spatial relations, maintaining probabilistic semantics while improving extrapolation to unseen primitive combinations. CoGe-GCD is implemented as an inductive-bias module between backbone and projection head, without modifying heads or losses, and can be plugged into diverse GCD frameworks. On standard benchmarks, it consistently improves all-class accuracy, unknown-class number estimation, and geometric quality, with marginal computational overhead. Code is available at https://github.com/lytang63/CoGe-GCD.
This report is categorized under AITOP's "Papers" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include CoGe-GCD, GCD, compositional generalization, computer vision.
Source attribution: curated from arXiv cs.LG (@Luyao Tang, Jiewei Zheng, Kunze Huang, Chaoqi Chen, Yue Huang, Cheng Chen).
