Data-DPO: Direct Preference Optimization for Target Model Data Selection in LLM Post-Training

arXiv:2608.16926v1 Announce Type: new Abstract: Data selection in supervised fine-tuning aims to select a small set of effective samples from large-scale candidate data, reducing training cost while preserving model performance. However, existing methods usually treat data value as a relatively static property, and pay limited attention to the compatibility between data and the capability distribution of the target model. To address this issue, we propose Data-DPO, a target model-oriented SFT data selection method. Data-DPO observes the local training feedback of the target model on different samples through one-step probing, transforms activation differences among samples into pairwise data preferences, and trains a lightweight reward model to learn target-model-aware data preferences. In the final selection stage, Data-DPO further combines target model preference, external quality scores, and marginal diversity to construct a more stable and effective training subset. Experimental results on Vision-Flan and LLaVA-CoT show that Data-DPO consistently outperforms existing data selection baselines under multiple data budgets and stably surpasses full data training performance.

Related

• The Long-Term Effects of Data Selection in LLM Fine-Tuning
• SEED: Targeted Data Selection by Weighted Independent Set
• Online Data Selection for Instruction Tuning via Gaussian Processes
• GLOBE: Trajectory-Aligned Gradient Matching with Structured SparseOptimization for Coreset Selection
• First-order Constrained Trilevel Optimization Over Distributed Networks for Robust Coreset Selection

Source: arXiv cs.LG | 2026-08-19