Score: 2

Extending the Context of Pretrained LLMs by Dropping Their Positional Embeddings

Published: December 13, 2025 | arXiv ID: 2512.12167v1

By: Yoav Gelberg , Koshi Eguchi , Takuya Akiba and more

Potential Business Impact:

Makes computers understand longer stories without retraining.

Business Areas:
Natural Language Processing Artificial Intelligence, Data and Analytics, Software

So far, expensive finetuning beyond the pretraining sequence length has been a requirement for effectively extending the context of language models (LM). In this work, we break this key bottleneck by Dropping the Positional Embeddings of LMs after training (DroPE). Our simple method is motivated by three key theoretical and empirical observations. First, positional embeddings (PEs) serve a crucial role during pretraining, providing an important inductive bias that significantly facilitates convergence. Second, over-reliance on this explicit positional information is also precisely what prevents test-time generalization to sequences of unseen length, even when using popular PE-scaling methods. Third, positional embeddings are not an inherent requirement of effective language modeling and can be safely removed after pretraining, following a short recalibration phase. Empirically, DroPE yields seamless zero-shot context extension without any long-context finetuning, quickly adapting pretrained LMs without compromising their capabilities in the original training context. Our findings hold across different models and dataset sizes, far outperforming previous specialized architectures and established rotary positional embedding scaling methods.

Country of Origin
🇬🇧 United Kingdom


Page Count
33 pages

Category
Computer Science:
Computation and Language