ColBERT Retrieval and Ensemble Response Scoring for Language Model Question Answering

Abstract: Domain-specific question answering remains challenging for LLMs, given the deep technical knowledge required to answer questions correctly. This difficulty is amplified for smaller LLMs that cannot encode as much information in their parameters as larger models. The "Specializing LLMs for Telecom Networks" challenge aimed to enhance the performance of two small LLMs, Phi-2 and Falcon-7B in telecommunication question answering. In this paper, we present our question answering systems for this challenge. Our solutions achieved leading marks of 81.9% accuracy for Phi-2 and 57.3% for Falcon-7B. We have publicly released our code and fine-tuned models.