Newest papers

TokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token Pruning

Liana Patel, Rishabh Ranjan, Ion Gabriel Stoica, Matei Zaharia, Carlos Guestrin · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26754 · cited by 0 works (OpenAlex) · open-access PDF

French

En revanche, les modèles linguistiques traitent tous les jetons d'entrée à travers chaque couche de modèle, ce qui entraîne des coûts significatifs de traitement pour les jetons de long contexte. Nous introduisons TokenGate, un petit composant drop-in qui élargit les modèles formés existants pour permettre la découpe de jetons apprenables de bout en bout sur les jetons de contexte fin pour des prélèvements rapides et efficaces en mémoire. Pour permettre l'apprentissage de bout en bout en dépit de l'opérateur Top-K non différenciable pour la découpe de jetons, nous introduisons un mécanisme inspiré par Mixture-of-Experts qui réévalue les jetons d'attention en utilisant des scores de mémoire

Machine translation (Meta M2M100); the English original follows.

Abstract

Expert human readers process large inputs (e.g., code repositories and long documents) by skimming to closely read only what matters. On the other hand, language models process all input tokens through each model layer, resulting in significant processing costs for long-context inputs. We introduce TokenGate, a small, drop-in component that extends existing trained models to enable end-to-end learnable token pruning over long-context inputs for fast and memory-efficient prefill. To enable end-to-end learning despite the non-differentiable top-K operator for token pruning, we introduce token-gated attention, a mechanism inspired by Mixture-of-Experts that re-weights attention logits using token importance scores. Across diverse tasks, TokenGate achieves $1.4$-$11.3\times$ faster time-to-first token (TTFT) and $1.5$-$3.4\times$ lower peak memory at no accuracy degradation for prefill in transformers. Compared to the best-performing prior methods for token-pruning, TokenGate gives up to $6.4\times$ faster prefill and $2\times$ lower peak memory. TokenGate natively extends both standard transformers (e.g., Qwen3) and hybrid-attention models (e.g., Qwen3.5) with simple lightweight adaptation comparable to standard parameter-efficient fine-tuning. We make our code available at: https://github.com/tokengate-models/TokenGate.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Security token, Computer science, Pruning, Artificial intelligence, Process (computing), Language model

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange