Gradient Ascent Post-training Enhances Language Model Generalization
This addresses the challenge of improving zero-shot generalization for language models, offering a promising method for broader NLP applications, though it appears incremental as it builds on existing post-training techniques.
The paper tackles the problem of enhancing language model generalization without task-specific fine-tuning by applying Gradient Ascent Post-training (GAP) on random, unlabeled text corpora, resulting in LMs becoming comparable to 2-3x larger models across 12 NLP tasks.
In this work, we empirically show that updating pretrained LMs (350M, 1.3B, 2.7B) with just a few steps of Gradient Ascent Post-training (GAP) on random, unlabeled text corpora enhances its zero-shot generalization capabilities across diverse NLP tasks. Specifically, we show that GAP can allow LMs to become comparable to 2-3x times larger LMs across 12 different NLP tasks. We also show that applying GAP on out-of-distribution corpora leads to the most reliable performance improvements. Our findings indicate that GAP can be a promising method for improving the generalization capability of LMs without any task-specific fine-tuning.