Modular LLM Architecture with Pluggable Reasoning Heads: A Scalable Approach to Multi-Modal AI Reasoning

Simple item page
dc.contributor.authorPathak, Anurag
dc.contributor.authorSharma, Dilip Kumar
dc.contributor.authorAgrawal, Harshada
dc.contributor.authorChawuthai, Rathachai
dc.contributor.authorPetchhan, Jirayu
dc.date.accessioned2026-08-06T10:48:41Z
dc.date.available2026-08-06T10:48:41Z
dc.date.issued2025-01-01
dc.description.abstractThis paper introduces a modular architecture for Large Language Models (LLMs) that incorporates pluggable, domain-specific reasoning heads to augment the model's capabilities beyond conventional text generation. Central to our approach is an attention-routing controller that intelligently classifies and dispatches user prompts to appropriate reasoning modules - namely symbolic, logical, or graph - based heads-based on the prompt's structure and intent. This design enables hybrid, multi-paradigm reasoning without requiring retraining or finetuning of the base LLM. By decoupling reasoning tasks from general language understanding, our system improves both computational efficiency and interpretability. We demonstrate the architecture using Groq as the base LLM and integrate lightweight engines such as SymPy for symbolic mathematics and custom modules for logical and graph-based reasoning. Experiments across a suite of structured and unstructured prompts show a significant reduction in inference latency and token usage, along with higher accuracy and better explainability. The proposed framework offers a scalable foundation for embedding modular reasoning capabilities into modern LLM-driven applications.
dc.identifier.citation2025 International Conference on Emerging Trends in Networks and Computer Communications Etncc 2025 Proceedings, 1099-1105, 2025
dc.identifier.doi10.1109/ETNCC66224.2025.11299615
dc.identifier.other2-s2.0-105031889899
dc.identifier.urihttps://dspace.kmitl.ac.th/handle/123456789/16318
dc.source2025 International Conference on Emerging Trends in Networks and Computer Communications Etncc 2025 Proceedings
dc.subjectExplainable AI
dc.subjectGroq
dc.subjectHybrid Models
dc.subjectLLMs
dc.subjectLogic Programming
dc.subjectModular AI
dc.subjectReasoning Heads
dc.subjectSymbolic AI
dc.subjectSymPy
dc.titleModular LLM Architecture with Pluggable Reasoning Heads: A Scalable Approach to Multi-Modal AI Reasoning
dc.typeConference Paper

Files

Collections

Research