Few-step text-to-image models increasingly replace slower generators, yet acceleration can silently change distributions over unspecified attributes even when individual outputs remain plausible and aligned. We call these distributions semantic defaults and their change under replacement semantic default shift. Existin...
Xuan-Hua Yin, Chuanzhi Xu, Shunqi Mao et al.· 2 citations
This work introduces ReAL, a training-free sampler that selects how far to advance using one shared lookahead, and shows that the discrepancy between uncorrected and lookahead-corrected endpoint proposals can be computed directly from the observed velocity mismatch and the candidate span beyond the lookahead.
Xuan-Hua Yin, Chuanzhi Xu, Hao-Xian Zhou et al.· 0 citations
PrefLUT is introduced, a reusable and refinable user-preference modeling framework for deployable 3D LUTs, encoding ordered preferred/non-preferred image pairs into a lightweight Reusable User Profile that is reused across queries and refined using additional user preference pairs, without per-user optimization.
Chuanzhi Xu, Lang-Yi Chen, Cheng-Kun Yue et al.· 0 citations
This work studies unpredictable near-optimal policies (UNOP), which mix uniformly on near-best actions that remain individually rational and which can be learned and the optimizer can match its best response, while the user's payoff stays higher because the mixture changes which action is targeted.
Zi-Han Deng, Chuanzhi Xu, Xiao-Zhen Zhong et al.· 0 citations
Aes3D, the first systematic framework for assessing the aesthetics of 3D neural rendering scenes, is proposed, and Aes3DGSNet, a lightweight model that directly predicts scene-level aesthetic scores from 3DGS representations, is presented, a lightweight model that operates solely on 3D Gaussian primitives.
This work proposes SciFigQual-Bench, a full-text contextual benchmark that evaluates scientific image across five dimensions (clarity, layout,captionfit, contextrelevance, andmisleadingrisk), and develops SFQ-Agent to achieve auditable and refined scoring through the collection and fusion of modal evidence.
SciFigAlign is proposed, a fine-tuned multimodal scorer that grounds figure quality assessment in manuscript evidence and confirms that manuscript-grounded inputs, citing-context denoising, and ranking supervision are all critical, showing that scientific figure assessment requires learned alignment between visual cont...
FracEvent, an event simulator that models this pixel-level lifecycle with fractional-relaxation voltage dynamics, improves the temporal structure of generated events and achieves stronger downstream-transfer results than competing simulator baselines, showing its practical value for event-camera simulation.
