Large language models (LLMs) now show strong reasoning and coding abilities, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. Existing benchmarks, however, mainly focus on isolated kernels, predefined operators, or pre-specified optimization targets, and...
This work introduces Universal Manipulation Representation (UMR), a unified action representation that enables zero-shot skill transfer from human demonstrations to heterogeneous robots and complements it with a Data-Efficient Strategy (DES) that diversifies object configurations through stage-aware point-cloud editing...
Song Liu, Lin-Ying Li, Yan-Shun Zhao et al.· arXiv.org· 0 citations
Denoising-Aware Visual Evidence Trajectory Allocation (DAVET), a training-free framework that allocates visual evidence according to the evolving generation state, achieves an average speedup of 1.55 times with an average relative performance drop of 1.86%, showing that denoising-aware visual evidence allocation can re...
Yongkang Zhou, Xiang-Wen Xia, Cheng Yan et al.· 0 citations
REFLEX is proposed, a training-free method that keeps the default router unchanged while reorganizing expert computation around the evolving refinement process, and introduces a coarse-to-fine hierarchy for expert-budget allocation that aligns computation with block-relative refinement roles while using the Frontier-Pr...
Xiang-Wen Xia, Chen Yan, Yiming Zhang et al.· 0 citations
RoleMerge, a training-free method that constructs each expert's Routing Role Profile (RRP) from phase-normalized routing statistics, capturing its relative phase preference, is proposed, and results validate phase-conditioned expert roles as a more effective basis than global routing aggregation for MoE-VLM expert merg...
Hong-Yu Zhang, Cheng Yan, Xiang-Wen Xia et al.· 1 citation
This work presents $\Phi$-Bench, a benchmark for systematically evaluating LLMs on engineering the LLM infrastructure stack and spans tasks of varying complexity, ranging from localized kernel-level function completion to long-horizon implementation and end-to-end system optimization.
Lei-Lei Ding, Shu-Min Wang, Yu-Ting Huang et al.· 1 citation
This framework is the first controlled, pre-registered instrument for this choice and never reads catch alone, and recovers up to 0.95 informedness over eight-action review, and no tested label-blind policy consistently beats it.