The paper introduces Dr. Post-Training, a framework that treats general training data as a data‑induced regularizer to prevent overfitting to scarce target objectives, rather than merely a pool for selection. It constructs a feasible set of model update directions from general data and projects target‑driven updates onto this set, unifying standard training and existing data‑selection methods as special cases along a bias‑variance spectrum. Experiments on SFT, RLHF, and RLVR demonstrate that these methods outperform both standard training and state‑of‑the‑art data‑selection baselines, especially when general and target data are highly misaligned, while system optimizations keep overhead minimal.
By Pingbang Hu, Xueshen Liu, Z. Morley Mao, Jiaqi W. Ma
arXiv:2610.14028v1 Announce Type: cross
Abstract: Data selection for LLM fine-tuning faces two facts about real training: data pools draw from many distinct domains, and evaluation spans diverse task...
By Luyang Zhang, Jialu Wang, Bebei Li
MetaJoin is a label‑free framework that adaptively selects join methods for heterogeneous tables by first choosing an initial specialist based on input evidence and then validating or repairing the result using specialist‑specific output evidence. It introduces MNN Join, a transformation specialist that finds string transformations by maximizing mutual‑nearest‑neighbor coverage and uses reciprocal validation. Evaluated on 670 datasets, MetaJoin achieves a macro row‑F1 of 0.817, surpassing a benchmark oracle and outperforming other label‑free policies by over 0.13, while MNN Join dominates all transformation baselines on 609 transformation‑heavy datasets.
By Ning Wang, Sainyam Galhotra
arXiv:2604. 26170v2 Announce Type: replace Abstract: Adapting large language models (LLMs) to a targeted task efficiently and effectively remains a fundamental challenge.
By Ting-Wei Li, Sirui Chen, Jiaru Zou, Yingbing Huang, Tianxin Wei, Jingrui He, Hanghang Tong
INCONVENIENT is a new benchmark that evaluates how machine‑learning models, especially large language models, generalize when trained on convenience samples that bias feature–target relationships. Unlike prior distribution‑shift tests that rely on natural domain splits, INCONVENIENT explicitly perturbs these relationships while keeping the deployment population unchanged, allowing controlled study of shift severity and direction. The benchmark covers 200 tabular classification and regression datasets and produces about 15,500 shifted training distributions, revealing significant performance degradation.
By Sara Rezaeimanesh, Chad Klochko, Kundan Thind, Mohammad Ghassemi
arXiv:2607. 20465v1 Announce Type: new Abstract: The quality of training data fundamentally determines the capabilities of large language models (LLMs), yet no unified benchmark exists to measure how well LLMs, agents, and data-centric workflows actually prepare training data end to end.
By Hao Liang, Qifeng Cai, Yibo Lin, Jianzhuo Du, Qifeng Xia, Sizhe Qiu, Linzhuang Sun, Meiyi Qiang, Zhaoyang Han, Xiaochen Ma, Bohan Zeng, Ruichuan An, Conghui He, Wentao Zhang
arXiv:2601.13111v3 Announce Type: replace-cross
Abstract: Realistic text-to-SQL workflows often require joining multiple tables. As a result, accurately retrieving the relevant set of tables becomes...
By Hassan Soliman, Vivek Gupta, Dan Roth, Iryna Gurevych
arXiv:2610.18572v1 Announce Type: new
Abstract: Automated post-training transforms pre-trained large language models (LLMs) into models capable of handling diverse domains by automatically constructi...
By Ruohong Liu, Bohan Tang, Lekang Jiang, Jack Umenberger, Yiwen Guo
arXiv:2602. 14696v2 Announce Type: replace Abstract: Instruction fine-tuning of large language models (LLMs) often involves selecting a subset of instruction training data from a large candidate pool, using a small query set from the target task.
By Nihal V. Nayak, Paula Rodriguez-Diaz, Neha Hulkund, Sara Beery, David Alvarez-Melis
The paper investigates how synthetic pretraining priors used in tabular foundation models (TFMs) influence downstream performance. By reconstructing the synthetic data generators of four TFMs and comparing their generated tasks to two popular tabular benchmarks using structural descriptors, the authors measure structural coverage and normalized density. They find that some generators provide broader and denser support for benchmark tasks, and that stronger synthetic-to-benchmark support generally correlates with better model performance.
By He Zhao, Ryan Thompson, Daniel M. Steinberg, Ashfaqur Rahman, Edwin V. Bonilla, Cheng Soon Ong
MetaSieve is a metapath selection layer that reduces subgraph size in relational deep learning by pruning uninformative metapaths using SQL join and aggregation statistics. It scores candidate metapath extensions with a lightweight function that favors informative yet lightweight paths, discarding those below a threshold. The method is independent of GNN parameters and, when applied to the RelBench benchmark, consistently cuts per‑epoch training time while preserving or improving accuracy.
By Fahim Shahriar Khan, Ashraf Aboulnaga
The paper proposes a fine‑tuning‑free listwise selector for Text‑to‑SQL systems that replaces traditional learning objectives with inference‑time strategies. It introduces reusable structured memories (MaP‑SQL) that encode mappings from natural language to schema elements, SQL operations, and expected outputs, and uses these memories to evaluate candidate queries. To reduce positional bias, the method aggregates rankings across multiple input permutations, optimizing inference cost through execution results and pointwise scoring. The approach achieves higher selection accuracy, fewer unnecessary comparisons, and outperforms the prior state‑of‑the‑art R^3‑SQL on the BIRD‑dev benchmark while using fewer tokens.
By Yeonseok Jeong, Soyoung Yoon, Seongjun Lee, Seung-won Hwang