Search papers, labs, and topics across Lattice.
This paper introduces Flash-BoN, a novel approach for inference-time scaling in text-to-image diffusion models that optimizes the generation process by generating a large pool of inexpensive draft candidates through timestep truncation, layer skipping, and activation proxies. The authors reveal that traditional methods, which focus on intermediate verification, often misrepresent efficiency, as simple Best-of-$N$ sampling can outperform more complex guided search techniques when evaluated under wall-clock time. Flash-BoN not only surpasses existing baselines across multiple benchmarks and model scales but also enhances the effectiveness of other techniques, demonstrating significant improvements in candidate diversity and RL post-training convergence.
Simple Best-of-$N$ sampling can outperform complex guided search methods in text-to-image generation, revealing a critical flaw in how we evaluate efficiency in diffusion models.
Inference-time scaling for text-to-image generation has progressed from simple Best-of-$N$ (BoN) sampling to guided search methods that verify and steer candidate trajectories at intermediate denoising steps. These approaches focus on when and how often to verify during denoising but largely treat the cost of generation itself as fixed. Moreover, the standard practice of comparing methods by number of function evaluations (NFEs) counts only denoising forward passes and ignores verifier overhead, which can distort efficiency rankings. We show that under wall-clock evaluation, simple BoN already matches or outperforms several guided search techniques, suggesting that compute is better spent on broader exploration than on repeated intermediate verification. This motivates Flash-BoN, which generates a large pool of inexpensive draft candidates by combining three complementary acceleration knobs: timestep truncation, layer skipping, and activation proxies into a single configuration optimized once per model. An efficient multi-stage verification procedure then identifies the most promising draft, which is refined at full quality. Across three benchmarks and three model scales, Flash-BoN consistently outperforms all baselines under fixed wall-clock budgets, with gains that grow at larger model scales (+8% AUC). We further show that our strategy combines well and improves existing orthogonal techniques such as reflection-based prompt optimization (+16% AUC). The gains correlate with increased candidate diversity, which also enables draft-guided selection to accelerate RL post-training convergence.
