ZooClaw-FashionSigLIP2：用於穩健時尚檢索的蒸餾微調

摘要

將基礎視覺語言編碼器應用於特定檢索任務時，會產生一個基本取捨：目標分佈上的效能提升，是以犧牲基礎模型的廣泛泛化能力為代價，而時尚檢索正是此問題的嚴峻案例。我們提出 ZooClaw-FashionSigLIP2，這是一個專注時尚領域的 SigLIP2-base 模型，透過一個簡單的策略解決了此取捨——在精心挑選的領域內數據上進行全微調搭配知識蒸餾，接著使用 \wiseft~wortsman2022wiseft 權重插值與基礎模型融合——其表現優於 LoRA、更大的骨幹模型（高達 1B 參數）以及外部訓練數據。在公平評估下，ZooClaw-FashionSigLIP2 在我們測試套件的所有基準上都優於所有基線。此外，我們釋出 ZooClaw-Fashion，一個新的高品質時尚檢索基準，以及對廣泛使用基準的系統性品質分析，該分析揭露並緩解了其公開真實標註中的結構性偏差。我們開源模型權重及所有評估工具，以促進未來研究。

English

Adapting a foundation vision-language encoder to a specialized retrieval task creates a fundamental tradeoff: gains on the target distribution come at the cost of the foundation model's broad generalization, and fashion retrieval is a stringent instance of this problem. We present ZooClaw-FashionSigLIP2, a fashion-specialized SigLIP2-base model that resolves this tradeoff with a simple recipe -- full fine-tuning with knowledge distillation on curated in-domain data, followed by \wiseft~wortsman2022wiseft weight interpolation with the base model -- and outperforms LoRA, larger backbones (up to 1B parameters), and external training data. Under fair evaluation, ZooClaw-FashionSigLIP2 outperforms all baselines on every benchmark in our suite. In addition, we release ZooClaw-Fashion, a new high-quality fashion retrieval benchmark, and a systematic quality analysis of widely-used benchmarks that exposes and mitigates structural biases in their public ground truth. We open-source the model weights and all evaluation artifacts to facilitate future research.