CorrectionLM: Self-Corrections with SLM for Dialogue State Tracking
This addresses the computational inefficiency of LLM-based corrections for SLMs in low-resource dialogue state tracking, though it is incremental as it builds on existing self-correction paradigms.
The paper tackles the problem of enabling small language models (SLMs) to self-correct without relying on large language models (LLMs), achieving results similar to a state-of-the-art LLM at a small fraction of the computation costs in dialogue state tracking tasks.
Large language models (LLMs) have demonstrated self-improvement capabilities via feedback and refinement, but current small language models (SLMs) have had limited success in this area. Existing correction approaches often rely on distilling knowledge from LLMs, which imposes significant computation demands. In this work, we introduce CORRECTIONLM, a novel correction framework that enables SLMs to self-correct using in-context exemplars without LLM involvement. Applied to two dialogue state tracking (DST) tasks in low-resource settings, CORRECTIONLM achieves results similar to a state-of-the-art LLM at a small fraction of the computation costs.