Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist

• URL: http://arxiv.org/abs/2606.31711v1
• Date: Tue, 30 Jun 2026 14:17:05 GMT
• Title: Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist
• Abstract summary: Existing faithfulness benchmarks rely on simple atomic instructions.<n>We introduce Arena-T2I Hard, a 310-prompt stress benchmark drawn from real arena T2I logs.<n>We propose a dependency-aware checklist reward that decomposes each prompt into a DAG of yes/no questions and zeroes descendants of failed parents.
• Score: 69.81453995715673
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Faithfulness -- how precisely a generated image aligns with its prompt -- is increasingly central to the real-world utility of text-to-image (T2I) models. Existing faithfulness benchmarks, however, rely on simple atomic instructions, on which top-tier systems already achieve near-perfect scores. As T2I models enter creative workflows, users issue multi-faceted requests combining intricate spatial relationships, stylistic constraints, and complex text rendering. In this setting, a single binary VLM-judge score no longer captures which specific constraints the model fails to satisfy. We introduce Arena-T2I Hard, a 310-prompt stress benchmark drawn from real arena T2I logs, with approximately 30 decomposed yes/no constraints per prompt spanning six categories, including text rendering. The strongest closed-source system we evaluate reaches 0.855 with a 33~pp performance gap across 11 systems, demonstrating substantial discriminative power. Moreover, high public-arena rankings fail to predict faithfulness, confirming that holistic Bradley-Terry (BT) preference scores prioritize aesthetics over fine-grained prompt adherence. We propose a dependency-aware checklist reward that decomposes each prompt into a DAG of yes/no questions and zeroes descendants of failed parents, turning faithfulness into a per-constraint training signal. Combined with a BT aesthetic reward via group-decoupled normalization (GDPO), which standardizes each reward within its rollout group so neither collapses, the recipe attains a strictly better faithfulness-aesthetics trade-off on SD3.5-Medium and FLUX.1-dev under MMRB2 pairwise comparisons than every single-reward, naive weighted-sum, or 4-reward BT-ensemble baseline.

Related papers

• OmniJudge or OmniBias? Diagnosing Multimodal Judges through Balanced, Decoupled Lenses [15.379006872674106]
  We introduce D3- Omni, a balanced and decoupled benchmark for diagnosing fine-grained multimodal understanding.<n>Rather than re-generating outputs, which may leak information across dimensions, we fix verified fully positive seeds and derive negatives.<n>The resulting D3 design is Dual-balanced, which helps alleviate negative-sample scarcity and per-dimension label imbalance.
  arXiv  Detail & Related papers  (2026-08-25T07:18:35Z)
• TAME: Test-Time Adversarial Prompt Tuning via Mixture-of-Experts for Vision-Language Models [76.92890872338235]
  Large-scale pre-trained Vision-Language models (VLMs) exhibit strong zero-shot generalization, yet are highly vulnerable to imperceptible adversarial perturbations.<n>To enhance robustness without requiring downstream task-specific retraining, we propose TAME, a novel test-time defense.
  arXiv  Detail & Related papers  (2026-05-17T18:07:08Z)
• DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models [78.62380562116135]
  We propose DynT2I-Eval, a fully automated dynamic evaluation framework for text-to-image (T2I) models.<n>It constructs a structured visual semantic space from long-form descriptions, decomposing prompts into controllable dimensions.<n>DynT2I-Eval evaluates model performance across text alignment, perceptual quality, and aesthetics.
  arXiv  Detail & Related papers  (2026-05-07T12:53:51Z)
• FIGROTD: A Friendly-to-Handle Dataset for Image Guided Retrieval with Optional Text [3.6723140587841656]
  Image-Guided Retrieval with Optional Text (IGROT) unifies visual retrieval (without text) and composed retrieval (with text)<n>We introduce FIGROTD, a lightweight yet high-quality IGROT dataset with 16,474 training triplets and 1,262 test triplets.<n>Trained on FIGROTD, VaGFeM achieves competitive results on nine benchmarks, reaching 34.8 mAP@10 on CIRCO and 75.7 mAP@200 on Sketchy.
  arXiv  Detail & Related papers  (2025-11-27T09:18:56Z)
• Easier Painting Than Thinking: Can Text-to-Image Models Set the Stage, but Not Direct the Play? [63.66192651248858]
  T2I-CoReBench is a comprehensive and complex benchmark that evaluates both composition and reasoning capabilities of T2I models.<n>To increase complexity, driven by the inherent real-world complexities, we curate each prompt with higher compositional density for composition.<n>In statistics, our benchmark comprises 1,080 challenging prompts and around 13,500 checklist questions.
  arXiv  Detail & Related papers  (2025-09-03T17:58:12Z)
• TIIF-Bench: How Does Your T2I Model Follow Your Instructions? [7.13169573900556]
  We present TIIF-Bench (Text-to-Image Instruction Following Benchmark), aiming to systematically assess T2I models' ability in interpreting and following intricate textual instructions.<n> TIIF-Bench comprises a set of 5000 prompts organized along multiple dimensions, which are categorized into three levels of difficulties and complexities.<n>Two critical attributes, i.e. text rendering and style control, are introduced to evaluate the precision of text synthesis and the aesthetic coherence of T2I models.
  arXiv  Detail & Related papers  (2025-06-02T18:44:07Z)
• Who Evaluates the Evaluations? Objectively Scoring Text-to-Image Prompt Coherence Metrics with T2IScoreScore (TS2) [62.44395685571094]
  We introduce T2IScoreScore, a curated set of semantic error graphs containing a prompt and a set of increasingly erroneous images. These allow us to rigorously judge whether a given prompt faithfulness metric can correctly order images with respect to their objective error count. We find that the state-of-the-art VLM-based metrics fail to significantly outperform simple (and supposedly worse) feature-based metrics like CLIPScore.
  arXiv  Detail & Related papers  (2024-04-05T17:57:16Z)
• Noisy-Correspondence Learning for Text-to-Image Person Re-identification [50.07634676709067]
  We propose a novel Robust Dual Embedding method (RDE) to learn robust visual-semantic associations even with noisy correspondences. Our method achieves state-of-the-art results both with and without synthetic noisy correspondences on three datasets.
  arXiv  Detail & Related papers  (2023-08-19T05:34:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.