Predicting Task Performance with Context-aware Scaling Laws
This work addresses the limitation of conventional scaling laws in capturing context-dependent performance, offering guidance for designing efficient long-context LLMs, though it is incremental as it builds on existing scaling law concepts.
The authors tackled the problem of predicting downstream task performance for large language models by developing a framework that incorporates context alongside training compute, achieving accurate modeling and generalization across three orders of magnitude in compute and reliable extrapolation with increased context.
Scaling laws have transformed our understanding of large language models by linking upstream metrics like cross-entropy loss to design factors such as model size, training data, and compute. However, these conventional laws fail to capture downstream task performance, where context plays a critical role. In this work, we propose a straightforward, interpretable framework that jointly models downstream performance as a function of the training compute and the provided context. We empirically validate our framework by fitting it on the observed downstream performance of extended-context variants of Llama-2-7B and Llama-2-13B across 65,500 unique instances spanning three tasks: arithmetic reasoning, common sense reasoning, and machine translation. Our results demonstrate that our framework accurately models in-distribution downstream performance, generalizes across three orders of magnitude in training compute, and reliably extrapolates performance as the amount of context increases. These findings offer valuable insights into the interplay between training compute and context utilization, providing guidance for designing more efficient long-context LLMs for diverse downstream tasks. Our code is available at https://github.com/wang-research-lab/context-scaling.