{"ID":2881441,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.12361","arxiv_id":"2508.12361","title":"Navigating the Exploration-Exploitation Tradeoff in Inference-Time Scaling of Diffusion Models","abstract":"Inference-time scaling has achieved remarkable success in language models, yet its adaptation to diffusion models remains underexplored. We observe that the efficacy of recent Sequential Monte Carlo (SMC)-based methods largely stems from globally fitting the The reward-tilted distribution, which inherently preserves diversity during multi-modal search. However, current applications of SMC to diffusion models face a fundamental dilemma: early-stage noise samples offer high potential for improvement but are difficult to evaluate accurately, whereas late-stage samples can be reliably assessed but are largely irreversible. To address this exploration-exploitation trade-off, we approach the problem from the perspective of the search algorithm and propose two strategies: Funnel Schedule and Adaptive Temperature. These simple yet effective methods are tailored to the unique generation dynamics and phase-transition behavior of diffusion models. By progressively reducing the number of maintained particles and down-weighting the influence of early-stage rewards, our methods significantly enhance sample quality without increasing the total number of Noise Function Evaluations. Experimental results on multiple benchmarks and state-of-the-art text-to-image diffusion models demonstrate that our approach outperforms previous baselines.","short_abstract":"Inference-time scaling has achieved remarkable success in language models, yet its adaptation to diffusion models remains underexplored. We observe that the efficacy of recent Sequential Monte Carlo (SMC)-based methods largely stems from globally fitting the The reward-tilted distribution, which inherently preserves di...","url_abs":"https://arxiv.org/abs/2508.12361","url_pdf":"https://arxiv.org/pdf/2508.12361v1","authors":"[\"Xun Su\",\"Jianming Huang\",\"Yang Yusen\",\"Zhongxi Fang\",\"Hiroyuki Kasai\"]","published":"2025-08-17T13:35:38Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.AI\",\"math.ST\"]","methods":"[\"Diffusion Model\",\"Language Model\",\"LoRA\"]","has_code":false}