Finer-Personalization Rank: Fine-Grained Retrieval Examines Identity Preservation for Personalized Generation

2025-12-21

AI Summary

Existing evaluation metrics for personalized generative models focus solely on semantic similarity and fail to assess identity preservation—particularly fine-grained, instance-level attributes (e.g., unique coat patterns in pets). Method: We propose the first retrieval-based, fine-grained evaluation protocol grounded in an identity-annotated gallery. It frames identity fidelity as a cross-image instance retrieval task, integrating fine-grained visual recognition with standard retrieval metrics (e.g., mean Average Precision), enabling multi-level assessment—from species and subclass to individual identity. Results: Experiments on CUB, Stanford Cars, and animal re-identification benchmarks demonstrate that our protocol effectively exposes significant identity drift in state-of-the-art methods, substantially outperforming conventional semantic metrics. It establishes a reproducible, decomposable, and identity-sensitive benchmark for personalized generation.

Technology Category

• Computer Vision: Image and Video Retrieval
• Natural Language Processing: Generation
• Knowledge Representation and Reasoning: Preferences

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements

Abstract

The rise of personalized generative models raises a central question: how should we evaluate identity preservation? Given a reference image (e.g., one's pet), we expect the generated image to retain precise details attached to the subject's identity. However, current generative evaluation metrics emphasize the overall semantic similarity between the reference and the output, and overlook these fine-grained discriminative details. We introduce Finer-Personalization Rank, an evaluation protocol tailored to identity preservation. Instead of pairwise similarity, Finer-Personalization Rank adopts a ranking view: it treats each generated image as a query against an identity-labeled gallery consisting of visually similar real images. Retrieval metrics (e.g., mean average precision) measure performance, where higher scores indicate that identity-specific details (e.g., a distinctive head spot) are preserved. We assess identity at multiple granularities -- from fine-grained categories (e.g., bird species, car models) to individual instances (e.g., re-identification). Across CUB, Stanford Cars, and animal Re-ID benchmarks, Finer-Personalization Rank more faithfully reflects identity retention than semantic-only metrics and reveals substantial identity drift in several popular personalization methods. These results position the gallery-based protocol as a principled and practical evaluation for personalized generation.

Problem

• Evaluating identity preservation in personalized generative models
• Addressing lack of fine-grained detail assessment in current metrics
• Introducing retrieval-based ranking for identity-specific detail evaluation

Innovation

• Ranking-based evaluation protocol for identity preservation
• Retrieval metrics measure fine-grained discriminative details
• Gallery-based assessment across multiple granularities and benchmarks