← Back to all news
arXiv Machine Learning By Lalchand Pandia, Kanishka Misra, Allyson Ettinger

Rethinking Data Selection: The Importance of Coverage over Difficulty in LLM Fine-Tuning

Read the original on arXiv Machine Learning →

The paper argues that for fine‑tuning large language models, selecting training data based on coverage of the input space is more effective than selecting based on difficulty. While difficulty‑based scores work for smaller classification models, they often underperform random sampling for LLMs. The authors propose a simple clustering‑based method that improves coverage and achieves performance equal to or better than random selection, especially on multi‑domain tasks.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv AI
Aug 26

Evaluating Multiple LLM Generations with Validated Task Coverage

The paper introduces VTC-Bench, a five‑domain benchmark designed to evaluate multiple outputs from large language models (LLMs) by measuring Validated Task Coverage (VTC). VTC quantifies how many distinct, useful results are produced within a set number of attempts, using real‑data tasks that allow automatic, reproducible checks of output quality and task‑relevant distinctness without relying on model‑based judges. Experiments show that models which perform best on single‑draw quality do not always achieve the highest coverage, and simple output‑variation metrics fail to capture task‑relevant diversity, highlighting the importance of evaluating finite candidate sets directly.

By Florian Le Bronnec, Rio Yokota
llmsbenchmarks
More like this →
arXiv Machine Learning
Aug 19

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

Data-DPO is a target model‑oriented supervised fine‑tuning data selection method that uses one‑step probing of the target model to generate pairwise data preferences, trains a lightweight reward model to capture these preferences, and then selects a training subset by combining target‑model preference, external quality scores, and marginal diversity. Experiments on Vision‑Flan and LLaVA‑CoT demonstrate that Data‑DPO consistently outperforms existing data selection baselines across multiple data budgets and even surpasses full data training performance.

By Peng Sun, Yi Yang, Antong Zhang, Chunxiao Li, Yanbo Wang, Dianbo Liu, xin chen, Kai Yu, Lu Chen, Tianfan Fu
llmsreinforcement-learningfine-tuning
More like this →