From Atoms to Chains: Divergence-Guided Reasoning Curriculum for Unlabeled LLM Domain Adaptation

2026-01-27

AI Summary

This work addresses the challenge of efficiently adapting large language models to specialized domains without labeled data, where existing knowledge distillation approaches often inherit the teacher model’s reasoning flaws due to coarse-grained imitation. To overcome this, the authors propose a Disagreement-Guided Reasoning Curriculum (DGRC) framework that automatically generates high-confidence atomic question-answer pairs by detecting discrepancies between teacher and student reasoning paths. DGRC constructs a dual-track dynamic curriculum that progresses from atomic knowledge to verified chains of thought, thereby relaxing the conventional distillation assumption of teacher infallibility and enabling unsupervised domain adaptation. Experiments in medical and legal domains demonstrate that a 1.5B-parameter student model trained with DGRC achieves a 7.76% relative accuracy improvement over strong baselines.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Knowledge Representation and Reasoning: Argumentation
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Adapting Large Language Models (LLMs) to specialized domains without human-annotated data is a crucial yet formidable challenge. Widely adopted knowledge distillation methods often devolve into coarse-grained mimicry, where the student model inefficiently targets its own weaknesses and risks inheriting the teacher's reasoning flaws. This exposes a critical pedagogical dilemma: how to devise a reliable curriculum when the teacher itself is not an infallible expert. Our work resolves this by capitalizing on a key insight: while LLMs may exhibit fallibility in complex, holistic reasoning, they often exhibit high fidelity on focused, atomic sub-problems. Based on this, we propose Divergence-Guided Reasoning Curriculum (DGRC), which constructs a learning path from atomic knowledge to reasoning chains by dynamically deriving two complementary curricula from disagreements in reasoning pathways. When a student and teacher produce conflicting results, DGRC directs the teacher to perform a diagnostic analysis: it analyzes both reasoning paths to formulate atomic queries that target the specific points of divergence, and then self-answers these queries to create high-confidence atomic question-answer pairs. These pairs then serve a dual purpose: (1) providing an atomic curriculum to rectify the student's knowledge gaps, and (2) serving as factual criteria to filter the teacher's original reasoning chains, yielding a verified CoT curriculum that teaches the student how to integrate atomic knowledge into complete reasoning paths. Experiments across the medical and legal domains on student models of various sizes demonstrate the effectiveness of our DGRC framework. Notably, our method achieves a 7.76% relative improvement for the 1.5B student model in the medical domain over strong unlabeled baseline.

Problem

• domain adaptation
• unlabeled data
• curriculum learning
• reasoning chains
• teacher-student disagreement

Innovation

• Divergence-Guided Reasoning Curriculum
• unlabeled domain adaptation
• atomic knowledge
• reasoning chains
• self-diagnostic distillation