pipette
ESEspañol

DeGMatch: Target-Aware Reasoning Data Selection via Deconfounded Gradient Matching

Changho Lee, Janghoon Han, Rajdeep Mukherjee, Yeda Song, Lening Nick Cui, Xinya Du, Honglak Lee

PreprintCode available

In the authors' words

Reasoning post-training on long trajectories has substantially improved the reasoning capabilities of large language models, but scaling such training to large datasets is increasingly expensive. We study target-aware reasoning data selection, where a small reference set specifies a target capability and guides the selection of training trajectories for specialization. We find that existing target-aware methods can exhibit limited target discrimination on long reasoning trajectories, while dominant population-level structure in candidate gradients can obscure finer target-dependent variation during gradient matching. To address these limitations, we introduce Deconfounded Gradient Matching (DeGMatch), which first constructs a target-relevance gate and then reranks candidates using model-specific gradients. DeGMatch centers gradients using the candidate-population mean and removes leading principal directions before target matching, reducing the influence of dominant population-level structure and improving target discrimination. Across mathematics, science, and code, DeGMatch outperforms strong heuristic, target-aware, and reasoning-specific data selection baselines, while remaining effective across selection budgets, model families, and scales. Our code is publicly available at https://github.com/CHLee0801/degmatch.

Main resultThe abstract does not state a limitation.

Appeared: Tuesday, October 20. arXiv. Preprint, not yet peer-reviewed.