paper-with-me

Papers

Data-Efficient Pretraining with Group-Level Data Influence Modeling

2025-02-20 · Zichun Yu, Fei Peng, Jie Lei, Arnold Overwijk, Wen-tau Yih, Chenyan Xiong

Data-efficient pretraining has shown tremendous potential to elevate scaling laws. This paper argues that effective pretraining data should be curated at the group level, treating a set of data points as a whole rather than as independent contributors. To achieve that, we propose Group-Level Data Influence Modeling (Group-MATES), a novel data-efficient pretraining method that captures and optimizes group-level data utility. Specifically, Group-MATES collects oracle group-level influences by locally probing the pretraining model with data sets. It then fine-tunes a relational data influence model to approximate oracles as relationship-weighted aggregations of individual influences. The fine-tuned model selects the data subset by maximizing its group-level influence prediction, with influence-aware clustering to enable efficient inference. Experiments on the DCLM benchmark demonstrate that Group-MATES achieves a 10% relative core score improvement on 22 downstream tasks over DCLM-Baseline and 5% over individual-influence-based methods, establishing a new state-of-the-art. Further analyses highlight the effectiveness of relational data influence models in capturing intricate interactions between data points.

📄 PDF Abstract BibTeX arXiv:2502.14709

Code (0)

등록된 구현이 없습니다.

Methods 이 논문이 사용한 방법론

SET Dynamic Sparse Training method where weight mask is updated randomly periodically

Similar Papers 제목 키워드 기반

Measuring Task-Agnostic Training Data Influence Across Language Model Pretraining

2026-08-13 · Yuto Nishida, Hirokazu Kiyomaru, Yusuke Oda, Takashi Kodama 외 arxiv

Measuring training data influence consistently across language model pretraining is challenging. It is difficult to select downstream tasks or validation sets representative of a model's general capabilities, and relianc…

Adversarial Pretraining of Self-Supervised Deep Networks: Past, Present and Future

2022-10-23 · Guo-Jun Qi, Mubarak Shah

In this paper, we review adversarial pretraining of self-supervised deep networks including both convolutional neural networks and vision transformers. Unlike the adversarial training with access to labeled examples, adv…

Contrastive LearningMiscellaneous

MATES: Model-Aware Data Selection for Efficient Pretraining with Data Influence Models

2024-06-10 · Zichun Yu, Spandan Das, Chenyan Xiong

Pretraining data selection has the potential to improve language model pretraining efficiency by utilizing higher-quality data from massive web data corpora. Current data selection methods, which rely on either hand-craf…

Language Modelling

DegAE: A New Pretraining Paradigm for Low-Level Vision

2023-01-01 · CVPR 2023 1 · Yihao Liu, Jingwen He, Jinjin Gu, Xiangtao Kong 외

Self-supervised pretraining has achieved remarkable success in high-level vision, but its application in low-level vision remains ambiguous and not well-established. What is the primitive intention of pretraining? Wh…

Philosophy

BLISS: A Lightweight Bilevel Influence Scoring Method for Data Selection in Language Model Pretraining

2025-10-07 · Jie Hao, Rui Yu, Wei Zhang, Huixia Wang 외 arxiv

Effective data selection is essential for pretraining large language models (LLMs), enhancing efficiency and improving generalization to downstream tasks. However, existing approaches often require leveraging external pr…

Bilevel Optimization