ALinFiK: Learning to Approximate Linearized Future Influence Kernel for Scalable Third-Parity LLM Data Valuation

2025-03-02

AI Summary

In large language model (LLM) training under constrained data budgets, existing data valuation methods suffer from prohibitive computational costs, poor scalability, and reliance on expensive second-order curvature estimation. Method: This paper introduces ALinFiK—the first third-party, scalable, single-sample data valuation framework for LLMs—built upon the Linearized Future Influence Kernel (LinFiK) paradigm. LinFiK integrates gradient linearization, influence function approximation, and meta-learning-driven kernel estimation, enabling real-time data value quantification for models with tens of millions of parameters. It avoids Hessian computation entirely, leveraging Hessian-free second-order optimization and efficient sampling strategies. Contribution/Results: On multiple LLM benchmarks, ALinFiK significantly outperforms state-of-the-art methods—including TracIn and Data Shapley—in both accuracy and efficiency: it achieves a 47× speedup in evaluation latency and processes over 200 million samples per day on a single GPU. This marks the first practical solution for high-accuracy, high-throughput, dynamic data valuation in large-scale LLM training.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large Language Models (LLMs) heavily rely on high-quality training data, making data valuation crucial for optimizing model performance, especially when working within a limited budget. In this work, we aim to offer a third-party data valuation approach that benefits both data providers and model developers. We introduce a linearized future influence kernel (LinFiK), which assesses the value of individual data samples in improving LLM performance during training. We further propose ALinFiK, a learning strategy to approximate LinFiK, enabling scalable data valuation. Our comprehensive evaluations demonstrate that this approach surpasses existing baselines in effectiveness and efficiency, demonstrating significant scalability advantages as LLM parameters increase.

Problem

• Develop scalable third-party data valuation for LLMs.
• Assess individual data sample impact on LLM performance.
• Propose ALinFiK for efficient and scalable data valuation.

Innovation

• Introduces linearized future influence kernel (LinFiK)
• Proposes ALinFiK for scalable data valuation
• Enhances LLM performance with efficient data assessment