Aligning Teacher with Student Preferences for Tailored Training Data Generation

TL;DR

ARTE introduces a novel teacher–student alignment framework to tailor training data for knowledge distillation in edge-ready LLMs. By combining knowledge elicitation, preference collection via in-context learning, and Direct Preference Optimization, ARTE produces task- and student-specific training examples that improve downstream reasoning on Big-Bench-Hard and out-of-domain benchmarks. Empirical results show ARTE outperforms existing instruction-tuning datasets and robustly generalizes across unseen tasks and student models with similar capacity. The work underscores the value of responsive teaching concepts in data-centric AI and suggests broader applicability to cross-task and cross-domain distillation scenarios.

Abstract

Large Language Models (LLMs) have shown significant promise as copilots in various tasks. Local deployment of LLMs on edge devices is necessary when handling privacy-sensitive data or latency-sensitive tasks. The computational constraints of such devices make direct deployment of powerful large-scale LLMs impractical, necessitating the Knowledge Distillation from large-scale models to lightweight models. Lots of work has been done to elicit diversity and quality training examples from LLMs, but little attention has been paid to aligning teacher instructional content based on student preferences, akin to "responsive teaching" in pedagogy. Thus, we propose ARTE, dubbed Aligning TeacheR with StudenT PreferencEs, a framework that aligns the teacher model with student preferences to generate tailored training examples for Knowledge Distillation. Specifically, we elicit draft questions and rationales from the teacher model, then collect student preferences on these questions and rationales using students' performance with in-context learning as a proxy, and finally align the teacher model with student preferences. In the end, we repeat the first step with the aligned teacher model to elicit tailored training examples for the student model on the target task. Extensive experiments on academic benchmarks demonstrate the superiority of ARTE over existing instruction-tuning datasets distilled from powerful LLMs. Moreover, we thoroughly investigate the generalization of ARTE, including the generalization of fine-tuned student models in reasoning ability and the generalization of aligned teacher models to generate tailored training data across tasks and students. In summary, our contributions lie in proposing a novel framework for tailored training example generation, demonstrating its efficacy in experiments, and investigating the generalization of both student & aligned teacher models in ARTE.

Figures (2)

• Figure 1: The overall framework of ARTE.
• Figure 2: The relationship between the word number of the rationale and the one-shot in-context learning accuracy of the Gemma-2B on boolean expressions and sports understanding tasks.