Arena-T2I Hard benchmark with ~30 decomposed constraints per prompt and a dependency-aware checklist reward yields better faithfulness-aesthetics trade-off than single-reward or weighted-sum baselines on SD3.5-Medium and FLUX.1-dev.
FLUX.2: Frontier Visual Intelligence
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
baseline 1
citation-polarity summary
years
2026 7roles
baseline 1polarities
baseline 1representative citing papers
BrainCause recovers known visual localizations and finds new candidate representations by validating causal specificity via counterfactual stimuli and encoding models, showing activation alone produces many false positives.
Edit-Compass and EditReward-Compass are new unified benchmarks for fine-grained image editing evaluation and realistic reward modeling in reinforcement learning optimization.
KVBench reveals major gaps in current T2I models for knowledge-intensive tasks, and KE-Check narrows the gap between open- and closed-source models by adding structured knowledge and enforcing constraints.
Projecting VAE latents to a fixed spherical radius and replacing linear interpolation with spherical linear interpolation improves class-conditional ImageNet-256 FID while leaving the diffusion architecture unchanged.
Massive activations form a sparse, spatially structured carrier subspace in DiTs that is necessary for generation quality and sufficient for training-free semantic transport across prompts and images.
VAE-LFA suppresses semantic drift in multi-turn DiT image editing by low-pass filtering latent discrepancies and aligning low-frequency components to an EMA of previous rounds in VAE space.
citing papers explorer
-
Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist
Arena-T2I Hard benchmark with ~30 decomposed constraints per prompt and a dependency-aware checklist reward yields better faithfulness-aesthetics trade-off than single-reward or weighted-sum baselines on SD3.5-Medium and FLUX.1-dev.
-
From Activation to Causality: Discovery of Causal Visual Representations in the Human Brain
BrainCause recovers known visual localizations and finds new candidate representations by validating causal specificity via counterfactual stimuli and encoding models, showing activation alone produces many false positives.
-
Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling
Edit-Compass and EditReward-Compass are new unified benchmarks for fine-grained image editing evaluation and realistic reward modeling in reinforcement learning optimization.
-
Knowledge Visualization: A Benchmark and Method for Knowledge-Intensive Text-to-Image Generation
KVBench reveals major gaps in current T2I models for knowledge-intensive tasks, and KE-Check narrows the gap between open- and closed-source models by adding structured knowledge and enforcing constraints.
-
Aligning Latent Geometry for Spherical Flow Matching in Image Generation
Projecting VAE latents to a fixed spherical radius and replacing linear interpolation with spherical linear interpolation improves class-conditional ImageNet-256 FID while leaving the diffusion architecture unchanged.
-
Few Channels Draw The Whole Picture: Revealing Massive Activations in Diffusion Transformers
Massive activations form a sparse, spatially structured carrier subspace in DiTs that is necessary for generation quality and sufficient for training-free semantic transport across prompts and images.
-
Why Do DiT Editors Drift? Plug-and-Play Low Frequency Alignment in VAE Latent Space
VAE-LFA suppresses semantic drift in multi-turn DiT image editing by low-pass filtering latent discrepancies and aligning low-frequency components to an EMA of previous rounds in VAE space.