VIEScore2: Unified Image Evaluation with Spatially Grounded Explanations
cs.CV
Submitted: 2026-10-01
Updated: 2026-10-01
Comments: Preprint. Project page: https://tiger-ai-lab.github.io/VIEScore2/
License: http://creativecommons.org/licenses/by/4.0/
The gist: Existing synthetic image evaluators typically provide only a scalar quality score and do not identify the image regions that support it.
Terminology
Abstract
Existing synthetic image evaluators typically provide only a scalar quality score and do not identify the image regions that support it. We introduce VIEScore2, a unified evaluator for image generation and editing tasks with optional conditioning images. VIEScore2 represents an image as an N x N grid and jointly predicts quality scores and defect locations in a single model pass. Its text-native grid representation provides a common interface for heterogeneous spatial supervision and enables directly verifiable post-training objectives. We train on 38K examples spanning score-only, localization-only, and joint supervision across generation and editing tasks. Starting from supervised fine-tuning, we further apply GRPO to improve defect localization using rewards that combine cell-level Dice overlap, score accuracy, and output-format validity. A parameter-free parser converts the structured predictions into readable explanations. On the primary suite, VIEScore2 achieves an overall-score SRCC of 0.601, compared with 0.491 for Gemini-3-Flash, the strongest zero-shot general-purpose VLM baseline under matched inputs. For defect localization, VIEScore2 outperforms both general-purpose VLMs and specialized spatial evaluators on three of six benchmarks in per-image grid IoU and ranks among the top three on five, including datasets beyond its training sources.
Sources
- Qwen3-VL Technical Report
- LoRA: Low-Rank Adaptation of Large Language Models
- Text4Seg: Reimagining Image Segmentation as Text Generation
- Evaluating Text-to-Visual Generation with Image-to-Text Generation
- Visual-RFT: Visual Reinforcement Fine-Tuning
- OmniQuality-R: Advancing Reward Models Through All-Encompassing Quality Assessment
- A Unified Agentic Framework for Evaluating Conditional Image Generation
- Detecting Human Artifacts from Text-to-Image Models
- Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis
- Where, What, Why, and Importance: Structured Defect Grounding for Text-to-Image Feedback
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models