Small Encoders Can Rival Large Decoders in Detecting Groundedness
This addresses the issue of ungrounded speculation in LLMs for NLP tasks, offering a resource-efficient solution for ensuring factual consistency and trustworthiness, though it is incremental as it builds on existing encoder models and datasets.
The study tackled the problem of detecting whether a query is grounded in provided context before costly answer generation by large language models, showing that lightweight encoder models like RoBERTa and NomicBERT achieve accuracy comparable to state-of-the-art LLMs such as Llama3 8B and GPT4o while reducing inference latency by orders of magnitude.
Augmenting large language models (LLMs) with external context significantly improves their performance in natural language processing (NLP) tasks. However, LLMs struggle to answer queries reliably when the provided context lacks information, often resorting to ungrounded speculation or internal knowledge. Groundedness - generating responses strictly supported by the context - is essential for ensuring factual consistency and trustworthiness. This study focuses on detecting whether a given query is grounded in a document provided in context before the costly answer generation by LLMs. Such a detection mechanism can significantly reduce both inference time and resource consumption. We show that lightweight, task specific encoder models such as RoBERTa and NomicBERT, fine-tuned on curated datasets, can achieve accuracy comparable to state-of-the-art LLMs, such as Llama3 8B and GPT4o, in groundedness detection while reducing inference latency by orders of magnitude. The code is available at : https://github.com/chandarlab/Hallucinate-less