AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Test-Time Alignment of Text-to-Image Diffusion Models via Null-Text Embedding Optimisation

Taehoon Kim, Henry Gouk, Timothy Hospedales

CVPR 2026 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
optimise null-text embedding (Null-TTA) instead of noise/latents, vs DNOSD-v1.5 · PickScore targetPickScore: 0.289 -> 0.315Improved
Null-TTA vs DAS on held-out ImageRewardSD-v1.5 · PickScore targetImageReward: 0.871 -> 0.946Improved
increase nmax from 25 to 115 (ablation)Null-TTA · HPSv2 targetHPSv2: 0.347 -> 0.428Improved
increase nmax from 25 to 115 (ablation)Null-TTA · HPSv2 targettime per image: 4m33s -> 17m07sHurt
optimise only null-text embedding rather than full networkNull-TTA · HPSv2 targetGPU memory (MB): 20449 -> 17585Improved
use Null-TTA instead of DNO at similar rewardSD-v1.5 · HPSv2 targetwall-clock time: 19m38s -> 8m40sImproved
apply Null-TTA to SDXLSDXL · PickScore targetPickScore: 0.253 -> 0.282Improved
increase nmax from 25 to 45 (ablation)Null-TTA · SDXL PickScore targetPickScore: 0.266 -> 0.282Improved
Null-TTA vs DNO on non-differentiable JPEG target (zeroth-order gradient)SD-v1.5 · JPEG compressibilityJPEG reward: -30.132 -> -33.741Mixed
Null-TTA vs DNO on held-out metrics under JPEG targetSD-v1.5 · JPEG compressibilityAesthetic: 3.988 -> 4.649Improved
Null-TTA vs DNO/DAS in user studySD-v1.5 · User study HPDv2mean rank: 2.04 -> 1.81Improved
Null-TTA vs DAS in multi-objective PickScore+HPSv2SD-v1.5 · Multi-objectivePareto frontierImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy