Reasoning Primitives in Hybrid and Non-Hybrid LLMs: Do Architectural Differences Yield Advantages in State-Tracking and Recall? 文章

ArXiv CS.CL2026-05-27NEWSen作者: Shivam Rawat, Lucie Flek, Florian Mai, Nicholas Kluge Corr\^ea
大语言模型
查看原文 →
关系图谱

Reasoning Primitives in Hybrid and Non-Hybrid LLMs: Do Architectural Differences Yield Advantages in State-Tracking and Recall? · 相关技术

相关技术

ginORMTTEODEPLALLMlanguage model企业通用语义标准divide-and-conquer partitioningaugmentationUCTTokenStraight-Through EstimatorStanSpatial Pivot-Aligned Coordinate-free Embedding (SPACE)SSESPAReferring expression comprehension (REC)MITLMMlarge language modelsHISGrouped Memorization EvaluationForARGANN