Can a Crow Hatch a Falcon? Lineage Matters in Predicting Large Language Model Performance
This work addresses the need for resource-efficient performance forecasting in LLM development, offering a novel approach to reduce computational costs and guide hyperparameter tuning, though it is incremental as it builds on existing matrix factorization methods.
The paper tackles the problem of predicting Large Language Model (LLM) performance before extensive fine-tuning or merging by proposing a Lineage-Regularized Matrix Factorization (LRMF) framework that incorporates lineage relationships, achieving up to 0.15-0.30 higher Pearson correlation coefficients compared to baseline methods in a study with 2,934 models and 21,000+ instances.
Accurately forecasting the performance of Large Language Models (LLMs) before extensive fine-tuning or merging can substantially reduce both computational expense and development time. Although prior approaches like scaling laws account for global factors such as parameter size or training tokens, they often overlook explicit lineage relationships-i.e., which models are derived or merged from which parents. In this work, we propose a novel Lineage-Regularized Matrix Factorization (LRMF) framework that encodes ancestral ties among LLMs via a graph Laplacian regularizer. By leveraging multi-hop parent-child connections, LRMF consistently outperforms conventional matrix factorization and collaborative filtering methods in both instance-level and benchmark-level performance prediction. Our large-scale study includes 2,934 publicly available Hugging Face models and 21,000+ instances across 6 major benchmarks, showing that the introduction of lineage constraints yields up to 0.15-0.30 higher Pearson correlation coefficients with actual performance compared to baseline methods. Moreover, LRMF effectively addresses the cold-start problem, providing accurate estimates for newly derived or merged models even with minimal data. This lineage-guided strategy thus offers a resource-efficient way to inform hyperparameter tuning, data selection, and model combination in modern LLM development.