Observational Scaling Laws and the Predictability of Language Model Performance

TL;DR

This paper introduces an observational approach to understanding language model performance scaling laws by analyzing ~100 publicly available models, revealing predictable phenomena and enabling performance forecasting without extensive training.

Contribution

It proposes a generalized scaling law based on a low-dimensional capability space, allowing prediction of model performance and emergent phenomena from existing models.

Findings

Performance follows a sigmoidal, predictable pattern.

Model capabilities can be forecasted from small models.

Post-training interventions' impacts can be predicted.

Abstract

Understanding how language model performance varies with scale is critical to benchmark and algorithm development. Scaling laws are one approach to building this understanding, but the requirement of training models across many different scales has limited their use. We propose an alternative, observational approach that bypasses model training and instead builds scaling laws from ~100 publically available models. Building a single scaling law from multiple model families is challenging due to large variations in their training compute efficiencies and capabilities. However, we show that these variations are consistent with a simple, generalized scaling law where language model performance is a function of a low-dimensional capability space, and model families only vary in their efficiency in converting training compute to capabilities. Using this approach, we show the surprising predictability of complex scaling phenomena: we show that several emergent phenomena follow a smooth, sigmoidal behavior and are predictable from small models; we show that the agent performance of models such as GPT-4 can be precisely predicted from simpler non-agentic benchmarks; and we show how to predict the impact of post-training interventions like Chain-of-Thought and Self-Consistency as language model capabilities continue to improve.