Skip to content
Preprint

T2LSC-Bench: Benchmarking Localized Semantic Control in Text-to-Image Generation

Sep 2026 · 1 citation · 43 references
Computer Science

TL;DR

A dual-branch protocol combines OCR-VLM text verification with structured VLM semantic judgments to measure Text-at-Anchor Accuracy (TAA), Semantic Subject Preservation (SSP), Semantic Leakage Rate (SLR), and Conditional Semantic Leakage Rate (cSLR).

Abstract

Recent text-to-image models have become increasingly capable of rendering explicit text, but reliable localized text control requires more than generating the correct string. In applications such as product labeling, signage, and interface design, target text should be rendered within a designated text-bearing region without altering the predefined subject identity or surrounding scene semantics. We refer to violations of this requirement as target-text-associated semantic leakage, in which target-text semantics are expressed through non-textual visual content beyond the designated anchor. Existing visual-text benchmarks primarily evaluate readability, spelling accuracy, and layout, leaving this form of semantic leakage largely unexamined. We introduce T2LSC-Bench, a controlled diagnostic benchmark comprising 50 seed subjects and 1,200 prompt cases per model, yielding 7,160 evaluated images across six models. Its factorized design varies semantic relation, scene openness, prompt mode, and language. A dual-branch protocol combines OCR-VLM text verification with structured VLM semantic judgments to measure Text-at-Anchor Accuracy (TAA), Semantic Subject Preservation (SSP), Semantic Leakage Rate (SLR), and Conditional Semantic Leakage Rate (cSLR). Under stress-test conditions, SLR increases from 1.2% to 18.1% and cSLR from 1.3% to 18.2%, whereas TAA decreases only from 91.4% to 90.9%. Anti-leakage prompting reduces SLR from 16.6% to 8.4% without degrading rendering accuracy. Human validation on 420 images shows strong agreement between automatic and adjudicated annotations. These results show that accurate text rendering does not guarantee local containment of target-text semantics.

View source

Similar papers

Book Open access Nov 2026

AGTI-Bench: A Human-Aligned Benchmark for Text-Aware Text-to-Image Generation

Rendering clear and accurate text is a critical capability of text-to-image (T2I) models. Despite rapid advances in T2I generation, the systematic and human-aligned evaluation of text rendering remains underexplored. Existing benchmarks often suffer from limited scenario coverage, a lack of bilingual support, and an ov...

Zhi-Chao Ma, Jia-Rui Wang, Xiong-Kuo Min · 0 citations
Preprint Oct 2026

TextVBench: Evaluating Text Rendering and Preservation in Video Generation

TextVBench is introduced, a benchmark and evaluation framework that formulates visual-text evaluation as a multi-target, multi-instance spatiotemporal correspondence problem, and is organized around controlled text, layout, carrier, and motion attributes so that aggregate scores can be traced back to specific capabilit...

Rui-Qiang Zhang, Heng-Yi Wang, Chang Liu et al. · 0 citations
Preprint Oct 2026

VTR-Bench: A Systematic Benchmark for Evaluating Visual Text Rendering in Video Generation

An automated evaluation pipeline with human alignments that separately assesses text fidelity through carrier-specific transcription and scene and motion requirements through a prompt-specific chain of query and a Director agent coordinates image and video generation with visual evaluation, guiding iterative refinement...

Yu Huang, Jun-Gang Li, Zhiyuan Wang et al. · 3 citations
Preprint Oct 2026

TextVerifier: Seeing Beyond Recognition to Verify Chinese Text Rendering

Text-to-image models increasingly generate readable Chinese text that still contains local rendering defects, such as a character with an extra or missing stroke or a malformed component. Such errors are invisible to OCR and vision-language models, which are trained for recognition and are therefore invariant to the ve...

Yi-Yang Chen, G. Wang, Yao-Li Liu et al. · 0 citations
#artificial intelligence Review Sep 2026

MatLoom: Layered Text-to-Material Generation in a Compact Program Space

Material generation should produce not only an appearance, but also the rules that construct it. We introduce MatLoom, a compact, layer-oriented language for text-to-material generation with pretrained language models. Each program composes alpha-masked layers whose shared spatial expressions define coverage and physic...

Anson Y. Lam, Shu-Qing Li, Michael R. Lyu · 0 citations
Conference Open access Sep 2026

Auxiliary text-guided image restoration for image-text matching

A new ITM framework that improves the model's discriminative performance by focusing on localized core attributes and has a better robustness in handling highly similar hard negatives, which provides a new way to address cross-modal hard sample discrimination.

Ya-Ting Liu, Kuang-Rong Hao · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.