← Back to Publications List

Cache-Aware Generative NLP: Reducing Latency in Large-Scale Multilingual Systems.

Students & Supervisors

Student Authors
Arnob Aich Anurag
Bachelor of Science in Computer Science & Engineering, FST
Md.nasif Rafidi
Bachelor of Science in Computer Science & Engineering, FST
Shamiul Islam
Bachelor of Science in Computer Science & Engineering, FST
Supervisors
Dr. Muhammad Firoz Mridha
Professor, Member [research & Publication Steering Committee], FST
Sultanul Arifeen Hamim
Lecturer, Faculty, FST

Abstract

Large-scale multilingual generative models face significant computational bottlenecks due to repetitive linguistic computations across languages. While existing caching mechanisms focus on token-level optimization, they overlook the computational redundancy in morphological and syntactic processing that occurs during multilingual text generation. We introduce Adaptive Linguistic Computation Caching (ALCC), a novel framework that identifies and caches expensive linguistic transformation computations during the generation process. ALCC employs a three-tier caching architecture: (1) Morphological Transformation Cache for expensive inflection computations, (2)Syntactic Pattern Cache for grammatical structure generation,and (3) Cross-Language Transfer Cache that leverages typological similarities to share computational patterns between related languages. Our approach introduces a Linguistic Complexity Predictor that dynamically identifies high-cost linguistic operations and prioritizes them for caching. Experiments on morphologically rich languages (Finnish, Turkish, Arabic) show an average of 75.8% latency reduction across 12 models on curated settings; gains vary with typological similarity and warm-up. in generation tasks while maintaining identical output quality. This work represents the first systematic approach to caching linguistic computations in generative multilingual systems.

Keywords

Natural Language Processing Multilingual Systems Caching Computational Efficiency Morphological Processing

Publication Details

  • Type of Publication:
  • Conference Name: 2026 IEEE 2nd International Conference on Quantum Photonics, Artificial Intelligence & Networking (QPAIN 2026)
  • Date of Conference: 16/04/2026 - 16/04/2026
  • Venue: IT Business Incubator, Chittagong University of Engineering and Technology (CUET), Chattogram, Bangladesh
  • Organizer: IEEE Photonics Society Bangladesh Chapter
View on Publisher