How Relevance Emerges: Interpreting LoRA Fine-Tuning in Reranking LLMs
This work addresses the need for mechanistic explanations of LoRA adaptation in information retrieval, offering incremental insights for researchers in that domain.
The paper tackled the problem of understanding how relevance signals are learned and deployed by large language models (LLMs) during LoRA fine-tuning for passage reranking, by fine-tuning models like Mistral-7B on MS MARCO and investigating factors such as LoRA rank and component importance, with results including insights into critical layers and projections for reranking accuracy.
We conduct a behavioral exploration of LoRA fine-tuned LLMs for Passage Reranking to understand how relevance signals are learned and deployed by Large Language Models. By fine-tuning Mistral-7B, LLaMA3.1-8B, and Pythia-6.9B on MS MARCO under diverse LoRA configurations, we investigate how relevance modeling evolves across checkpoints, the impact of LoRA rank (1, 2, 8, 32), and the relative importance of updated MHA vs. MLP components. Our ablations reveal which layers and projections within LoRA transformations are most critical for reranking accuracy. These findings offer fresh explanations into LoRA's adaptation mechanisms, setting the stage for deeper mechanistic studies in Information Retrieval. All models used in this study have been shared.