Caption Bottleneck Models
19th European Conference on Computer Vision, ECCV 2026, Malmö, İsveç, 8 - 12 Eylül 2026, cilt.17036 LNCS, ss.445-461, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 17036 LNCS
- Doi Numarası: 10.1007/978-3-032-37132-4_25
- Basıldığı Şehir: Malmö
- Basıldığı Ülke: İsveç
- Sayfa Sayıları: ss.445-461
- Anahtar Kelimeler: Concept Bottleneck Models, Concept Discovery
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
Concept Bottleneck Models (CBMs) provide interpretability by routing predictions through a layer of human-understandable concepts. However, defining an optimal concept set for a specific dataset remains an open challenge. Existing approaches rely on expensive expert annotations or LLM-generated lists based solely on class names. Even “open-vocabulary” variants typically depend on static concept sets, which restrict discovery and introduce label bias. Furthermore, traditional CBMs often suffer from information leakage, where unmodeled visual features bypass the bottleneck and compromise the integrity of the explanations. To overcome these limitations, we propose Caption Bottleneck Models (CaBM), a framework that circumvents the need for predefined concept sets by replacing rigid concept layers with free-form natural language. By representing images via LMM-generated captions and training a classifier strictly on this text, CaBM ensures a leakage-free architecture by construction. Additionally, by analyzing the text classifier post-training, CaBM autonomously discovers high-quality, dataset-specific concepts. Our results across fine- and coarse-grained benchmarks demonstrate that CaBM achieves competitive accuracy while preserving interpretability without the constraints of external dictionaries or manual labeling. Our code is available at https://github.com/bariscagliyan/CaptionBottleneckModels.