SMART SLM: Structured Memory and Reasoning Transformer, A Small Language Model for Accurate Document Assistance
Divij Dudeja, Mayukha Pal
TL;DR
SMART addresses the challenge of exact, fast QA over long, structured engineering manuals by separating extraction, memory, and reasoning. A Grammarian (Tree-LSTM) converts sentences into canonical subject–relation–object facts, stored in a Memory Librarian as fixed-size vectors with provenance, while a 6-layer Transformer fuses retrieved facts through gated memory attention. At 45.51M parameters, SMART is significantly more parameter-efficient than GPT-2 and BERT, while achieving higher accuracy and reducing hallucinations via an explicit factual store and dual-path inference (pre-indexed for known documents and RAG-assisted for new uploads). This design yields auditable, provenance-backed answers with reduced latency and demonstrates a practical blueprint for reliable document-aware AI in technical domains.
Abstract
The user of Engineering Manuals (EM) finds it difficult to read EM s because they are long, have a dense format which includes written documents, step by step procedures, and standard parameter lists for engineering equipment. Off the shelf transformers, especially compact ones, treat this material as a flat stream of tokens. This approach leads to confident but incorrect numeric answers and forces the models to memorize separate facts inefficiently. SMART (Structured Memory and Reasoning Transformer) offers a different and practical solution to the above problem. SMART structures its processing by using a hierarchical approach, and is based upon three main job categories (1) A syntax-aware Fact Extractor (Grammarian) Tree LSTM which extracts facts as subject relation object relations from EM sentences (2) A compact indexed memory MANN (Memory Augmented Neural Network) that indexes these Rational Subject Relation Objects as 384 dimensional vectors that are associated with the source of the information, and (3) A 6 layer Transformer that learns to fuse the previously retrieved facts into its generated response. The entire SMART model utilizes 45.51M parameters, which is 64% less than GPT-2 (124M) and 69% less than BERT (133M), and it achieves a 21.3% higher accuracy than GPT-2, indicating that SMART fits the data better with the least amount of processing requirements. SMART employs dual modes of inference an indexed fast path for known documents (sub-second answer times) and an indexed dynamic path assisted by RAGs for new uploads (FAISS Top 20 results with memory severed at 64 slots). In real world deployment, this framework leads to more well supported results with reduced hallucinations than comparable small transformer models.