[2604.03428] Inference-Path Optimization via Circuit Duplication in Frozen Visual Transformers for Marine Species Classification

Computer Science > Computer Vision and Pattern Recognition arXiv:2604.03428 (cs) [Submitted on 3 Apr 2026] Title:Inference-Path Optimization via Circuit Duplication in Frozen Visual Transformers for Marine Species Classification Authors:Thomas Manuel Rost View a PDF of the paper titled Inference-Path Optimization via Circuit Duplication in Frozen Visual Transformers for Marine Species Classification, by Thomas Manuel Rost View PDF HTML (experimental) Abstract:Automated underwater species classification is constrained by annotation cost and environmental variation that limits the transferability of fully supervised models. Recent work has shown that frozen embeddings from self-supervised vision foundation models already provide a strong label-efficient baseline for marine image classification. Here we investigate whether this frozen-embedding regime can be improved at inference time, without fine-tuning or changing model weights. We apply Circuit Duplication, an inference-time method originally proposed for Large Language Models, in which a selected range of transformer layers is traversed twice during the forward pass. We evaluate on the class-imbalanced AQUA20 benchmark using frozen DINOv3 embeddings under two settings: global circuit selection, where a single duplicated circuit is chosen for the full dataset, and class-specific circuit selection, where each species may receive a different optimal circuit. Both settings use simple semi-supervised downstream classifiers. C...