Search papers, labs, and topics across Lattice.
This paper introduces DOMINO, a novel framework for domain-specific data synthesis that shifts from a deductive to an inductive paradigm, using only reference examples to define target domains. By integrating prompt tuning with a contrastive disentanglement objective, DOMINO effectively learns minimal sufficient domain representations, allowing for the generation of diverse, domain-aligned synthetic data while mitigating overfitting. Empirical results show that fine-tuning on data synthesized by DOMINO significantly improves Pass@1 accuracy on challenging coding benchmarks, outperforming strong instruction-tuned models by up to 4.63%.
Shifting from explicit domain descriptions to an inductive approach, DOMINO generates high-quality synthetic data that enhances LLM performance without the need for manual prompt engineering.
Large Language Models have demonstrated remarkable progress in general-purpose capabilities and can achieve strong performance in specific domains through fine-tuning on domain-specific data. However, acquiring high-quality data for target domains remains a significant challenge. Existing data synthesis approaches follow a deductive paradigm, heavily relying on explicit domain descriptions expressed in natural language and careful prompt engineering, limiting their applicability in real-world scenarios where domains are difficult to describe or formally articulate. In this work, we tackle the underexplored problem of domain-specific data synthesis through an inductive paradigm, where the target domain is defined only through a set of reference examples, particularly when domain characteristics are difficult to articulate in natural language. We propose a novel framework, DOMINO, that learns a minimal sufficient domain representation from reference samples and leverages it to guide the generation of domain-aligned synthetic data. DOMINO integrates prompt tuning with a contrastive disentanglement objective to separate domain-level patterns from sample-specific noise, mitigating overfitting while preserving core domain characteristics. Theoretically, we prove that DOMINO expands the support of the synthetic data distribution, ensuring greater diversity. Empirically, on challenging coding benchmarks where domain definitions are implicit, fine-tuning on data synthesized by DOMINO improves Pass@1 accuracy by up to 4.63\% over strong, instruction-tuned backbones, demonstrating its effectiveness and robustness. This work establishes a new paradigm for domain-specific data synthesis, enabling practical and scalable domain adaptation without manual prompt design or natural language domain specifications.
