Learning to Adapt SFT Data for Better Reasoning Generalization

• URL: http://arxiv.org/abs/2605.26924v1
• Date: Tue, 26 May 2026 12:20:53 GMT
• Title: Learning to Adapt SFT Data for Better Reasoning Generalization
• Abstract summary: supervised fine-tuning (SFT) is widely used to provide dense supervision and enables efficient training.<n>Data Adaptation for Reasoning Tuning (DART) trains a mapper model with reinforcement learning to convert SFT data into model-adapted supervision.<n>Experiments show that DART improves generalization, achieves higher training efficiency than direct RL, and helps models surpass standard SFT.
• Score: 14.088766743760575
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) have achieved remarkable progress, with post-training playing a crucial role in enhancing their reasoning capabilities. Among post-training paradigms, supervised fine-tuning (SFT) is widely used: it leverages external data to provide dense supervision and enables efficient training. However, directly fine-tuning on expert data can hurt generalization when the data distribution is mismatched with the target model's own distribution. In this work, we propose Data Adaptation for Reasoning Tuning (DART), which formulates the use of a fixed, potentially distributionally misaligned SFT dataset as an optimization problem over demonstration transformations. DART trains a mapper model with reinforcement learning to convert original SFT data into model-adapted supervision that better matches the target model's distribution and learning preferences. The transformed data are then used for SFT, allowing the target model to better exploit external supervision. Experiments across multiple models and datasets show that DART improves generalization, achieves higher training efficiency than direct RL, and helps models surpass standard SFT. Our code is available at https://anonymous.4open.science/r/DART525E50D.

Related papers

• Theoretical Perspectives on Data Quality and Synergistic Effects in Pre- and Post-Training Reasoning Models [56.12341509545198]
  Large Language Models (LLMs) are pretrained on massive datasets and later instruction-tuned via supervised fine-tuning (SFT) or reinforcement learning (RL)<n>Best practices emphasize large, diverse pretraining data, whereas post-training operates differently.<n>We theoretically analyze transformers trained on an in-context weight prediction task for linear regression.
  arXiv  Detail & Related papers  (2026-03-01T21:58:09Z)
• Learning from the Undesirable: Robust Adaptation of Language Models without Forgetting [18.680059467974825]
  Language models (LMs) are often adapted through supervised fine-tuning (SFT) to specialize their capabilities for downstream tasks.<n>In typical scenarios where the fine-tuning data is limited, SFT can lead LMs to overfit, causing them to rely on spurious patterns.<n>We propose Learning-from-the-Undesirable (LfU), a simple yet effective regularization scheme for SFT to mitigate issues when fine-tuning LMs with limited data.
  arXiv  Detail & Related papers  (2025-11-17T06:57:44Z)
• The Harder The Better: Maintaining Supervised Fine-tuning Generalization with Less but Harder Data [6.136716058442803]
  We propose THTB (The Harder The Better), a cognitive science-inspired framework for instruction data selection and annotation guidance.<n>Experiments show THTB enables models trained on only 5% of the data to outperform full-dataset training.<n>In addition, THTB provides effective annotation guidance in vertical domains, enabling a model trained on just 2% of the data to surpass models trained on much larger datasets.
  arXiv  Detail & Related papers  (2025-10-14T08:25:24Z)
• InfiAlign: A Scalable and Sample-Efficient Framework for Aligning LLMs to Enhance Reasoning Capabilities [27.09178257629886]
  InfiAlign is a scalable and sample-efficient post-training framework for large language models (LLMs)<n>At the core of InfiAlign is a robust data selection pipeline that automatically curates high-quality alignment data from open-source reasoning.<n>Our results highlight the effectiveness of combining principled data selection with full-stage post-training.
  arXiv  Detail & Related papers  (2025-08-07T15:34:06Z)
• Improved Supervised Fine-Tuning for Large Language Models to Mitigate Catastrophic Forgetting [1.5595148909011116]
  Supervised Fine-Tuning (SFT) is a critical step for enhancing the instruction-following capabilities of Large Language Models (LLMs)<n>SFT often leads to a degradation of the model's general abilities, a phenomenon known as catastrophic forgetting.<n>We propose a novel and cost-effective SFT method that effectively mitigates catastrophic forgetting without requiring access to the original SFT data.
  arXiv  Detail & Related papers  (2025-06-11T06:23:50Z)
• OpenCodeReasoning: Advancing Data Distillation for Competitive Coding [61.15402517835137]
  We build a supervised fine-tuning (SFT) dataset to achieve state-of-the-art coding capability results in models of various sizes.<n>Our models use only SFT to achieve 61.8% on LiveCodeBench and 24.6% on CodeContests, surpassing alternatives trained with reinforcement learning.
  arXiv  Detail & Related papers  (2025-04-02T17:50:31Z)
• Discriminative Finetuning of Generative Large Language Models without Reward Models and Human Preference Data [73.04828796123581]
  Supervised fine-tuning (SFT) has become a crucial step for aligning pretrained large language models (LLMs)<n>We introduce Discriminative Fine-Tuning (DFT), an improved variant of SFT, which mitigates the burden of collecting human-labeled preference data.<n>Our contributions include: (i) a discriminative probabilistic framework for fine-tuning LLMs by explicitly modeling the discriminative likelihood of an answer among all possible outputs given an input; (ii) efficient algorithms to optimize this discriminative likelihood; and (iii) extensive experiments demonstrating DFT's effectiveness
  arXiv  Detail & Related papers  (2025-02-25T22:38:55Z)
• Preference-Oriented Supervised Fine-Tuning: Favoring Target Model Over Aligned Large Language Models [12.500777267361102]
  We introduce a novel textbfpreference-textbforiented supervised textbffine-textbftuning approach, namely PoFT.<n>The intuition is to boost SFT by imposing a particular preference: textitfavoring the target model over aligned LLMs on the same SFT data.<n>PoFT achieves stable and consistent improvements over the SFT baselines across different training datasets and base models.
  arXiv  Detail & Related papers  (2024-12-17T12:49:14Z)
• Adaptive Data Optimization: Dynamic Sample Selection with Scaling Laws [59.03420759554073]
  We introduce Adaptive Data Optimization (ADO), an algorithm that optimize data distributions in an online fashion, concurrent with model training. ADO does not require external knowledge, proxy models, or modifications to the model update. ADO uses per-domain scaling laws to estimate the learning potential of each domain during training and adjusts the data mixture accordingly.
  arXiv  Detail & Related papers  (2024-10-15T17:47:44Z)
• Functional Graphical Models: Structure Enables Offline Data-Driven Optimization [111.28605744661638]
  We show how structure can enable sample-efficient data-driven optimization. We also present a data-driven optimization algorithm that infers the FGM structure itself.
  arXiv  Detail & Related papers  (2024-01-08T22:33:14Z)
• DavIR: Data Selection via Implicit Reward for Large Language Models [62.59514469369608]
  DavIR is a model-based data selection method for post-training Large Language Models.<n>We show that 6% of Alpaca dataset selected with DavIR can steer both the LLaMA and Gemma model family to produce superior performance compared to the same models trained on the full 52K dataset.
  arXiv  Detail & Related papers  (2023-10-16T07:26:24Z)
• Scaling Relationship on Learning Mathematical Reasoning with Large Language Models [75.29595679428105]
  We investigate how the pre-training loss, supervised data amount, and augmented data amount influence the reasoning performances of a supervised LLM. We find that rejection samples from multiple models push LLaMA-7B to an accuracy of 49.3% on GSM8K which outperforms the supervised fine-tuning (SFT) accuracy of 35.9% significantly.
  arXiv  Detail & Related papers  (2023-08-03T15:34:01Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.