提升扩散多模态大模型生成质量与效率，自验证式测试时扩展新方法。
dMLLM-TTS: Self-Verified and Efficient Test-Time Scaling for Diffusion Multi-Modal Large Language Models
- 分两轴并行探索轨迹与迭代优化，提升生成多样性与稳定性。
- 在GenEval上性能显著优于线性搜索，效率最高提升6倍。
- 无需外部验证器，利用模型自身理解能力实现自验证。
扩散多模态大语言模型（dMLLMs）近期成为融合图像生成与理解的新架构。然而，如何高效实现测试时扩展（TTS）以释放其全部生成潜力仍是未充分探索的挑战。为此，我们提出dMLLM-TTS框架，基于两个互补扩展维度：(1) 轨迹探索扩展以增强生成假设多样性，(2) 迭代精炼扩展以保证生成稳定性。传统TTS方法通常在线性搜索两个维度上进行，计算复杂度达O(NT)，且需外部验证器完成最佳N选一。为克服此限制，我们提出两项创新：首先设计一种高效分层搜索算法，复杂度降至O(N+T)，可自适应扩展与剪枝采样轨迹；其次引入自验证反馈机制，利用dMLLM内在图像理解能力评估图文对齐，无需外部验证器。在GenEval基准上对三种代表性dMLLM（如Lumina-DiMOO、MMaDA、Muddit）的大量实验表明，本框架显著提升生成质量，同时相较线性搜索最高实现6倍效率提升。项目页：https://github.com/Alpha-VLLM/Lumina-DiMOO。
原文摘要 · Abstract (English)
Diffusion Multi-modal Large Language Models (dMLLMs) have recently emerged as a novel architecture unifying image generation and understanding. However, developing effective and efficient Test-Time Scaling (TTS) methods to unlock their full generative potential remains an underexplored challenge. To address this, we propose dMLLM-TTS, a novel framework operating on two complementary scaling axes: (1) trajectory exploration scaling to enhance the diversity of generated hypotheses, and (2) iterative refinement scaling for stable generation. Conventional TTS approaches typically perform linear search across these two dimensions, incurring substantial computational costs of O(NT) and requiring an external verifier for best-of-N selection. To overcome these limitations, we propose two innovations. First, we design an efficient hierarchical search algorithm with O(N+T) complexity that adaptively expands and prunes sampling trajectories. Second, we introduce a self-verified feedback mechanism that leverages the dMLLMs' intrinsic image understanding capabilities to assess text-image alignment, eliminating the need for external verifier. Extensive experiments on the GenEval benchmark across three representative dMLLMs (e.g., Lumina-DiMOO, MMaDA, Muddit) show that our framework substantially improves generation quality while achieving up to 6x greater efficiency than linear search. Project page: https://github.com/Alpha-VLLM/Lumina-DiMOO.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。