The Super Weight in Large Language Models

TL;DR

This paper reveals that a tiny subset of parameters, called super weights, are crucial for LLM performance, and removing or quantizing them significantly impairs the model, leading to new methods for identifying and preserving these parameters.

Contribution

The paper introduces a data-free method to identify super weights in LLMs, demonstrating their critical role and proposing improved quantization techniques by preserving these weights.

Findings

Removing a single super weight drastically increases perplexity.

Preserving super weights improves quantization accuracy.

Super weights induce large activation outliers.

Abstract

Recent works have shown a surprising result: a small fraction of Large Language Model (LLM) parameter outliers are disproportionately important to the quality of the model. LLMs contain billions of parameters, so these small fractions, such as 0.01%, translate to hundreds of thousands of parameters. In this work, we present an even more surprising finding: Pruning as few as a single parameter can destroy an LLM's ability to generate text -- increasing perplexity by 3 orders of magnitude and reducing zero-shot accuracy to guessing. We propose a data-free method for identifying such parameters, termed super weights, using a single forward pass through the model. We additionally find that these super weights induce correspondingly rare and large activation outliers, termed super activations. When preserved with high precision, super activations can improve simple round-to-nearest quantization to become competitive with state-of-the-art methods. For weight quantization, we similarly find that by preserving the super weight and clipping other weight outliers, round-to-nearest quantization can scale to much larger block sizes than previously considered. To facilitate further research into super weights, we provide an index of super weight coordinates for common, openly available LLMs.