Seeking Neural Nuggets: Knowledge Transfer in Large Language Models from a Parametric Perspective
This work addresses the challenge of knowledge transfer across models of varying scales for AI practitioners, but it is incremental as it builds on existing sensitivity-based and LoRA techniques.
The paper tackles the problem of transferring knowledge from larger to smaller language models by extracting and aligning knowledge-specific parameters, achieving validated efficacy across four benchmarks.
Large Language Models (LLMs) inherently encode a wealth of knowledge within their parameters through pre-training on extensive corpora. While prior research has delved into operations on these parameters to manipulate the underlying implicit knowledge (encompassing detection, editing, and merging), there remains an ambiguous understanding regarding their transferability across models with varying scales. In this paper, we seek to empirically investigate knowledge transfer from larger to smaller models through a parametric perspective. To achieve this, we employ sensitivity-based techniques to extract and align knowledge-specific parameters between different LLMs. Moreover, the LoRA module is used as the intermediary mechanism for injecting the extracted knowledge into smaller models. Evaluations across four benchmarks validate the efficacy of our proposed method. Our findings highlight the critical factors contributing to the process of parametric knowledge transfer, underscoring the transferability of model parameters across LLMs of different scales. Project website: https://maszhongming.github.io/ParaKnowTransfer.