Random Scaling of Emergent Capabilities

TL;DR

This paper investigates how emergent capabilities in language models are driven by continuous changes in training outcome distributions across random seeds, rather than abrupt threshold effects, revealing the importance of seed variability in performance scaling.

Contribution

It demonstrates that emergent capabilities result from gradual distribution shifts across seeds, challenging the notion of sudden threshold-based breakthroughs in model scaling.

Findings

Emergent performance is linked to bimodal distribution shifts across seeds.

Sharp metric breakthroughs are due to underlying distribution changes, not thresholds.

Seed variability significantly influences performance predictions at scale.

Abstract

Language models famously improve under a smooth scaling law, but some specific capabilities exhibit sudden breakthroughs in performance. Advocates of "emergence" view these capabilities as unlocked at a specific scale, but others attribute breakthroughs to superficial metric thresholding effects. We propose that breakthroughs are instead driven by continuous changes in the probability distribution of training outcomes when performance is bimodally distributed across random seeds. we show that different random seeds can produce either smooth or emergent scaling trends in synthetic length generalization tasks, multiple choice question answering, and grammatical generalization. We reveal that sharp breakthroughs in metrics are produced by underlying continuous changes in their distribution across seeds. These distributions may become abruptly bimodal at a capacity threshold but this threshold appears at scales well before most seeds achieve breakthrough. Our observations hold true even under continuous loss metrics, confirming that random variation must be considered when predicting a model's performance from its scale.