Salvato in:
Dettagli Bibliografici
Autori principali: Silva, Pedro Luiz, de Domenico, Antonio, Maatouk, Ali, Ayed, Fadhel
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2409.19001
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913521787207680
author Silva, Pedro Luiz
de Domenico, Antonio
Maatouk, Ali
Ayed, Fadhel
author_facet Silva, Pedro Luiz
de Domenico, Antonio
Maatouk, Ali
Ayed, Fadhel
contents Despite the remarkable success of Large Language Models (LLMs), they still exhibit a limited capability to align their outputs to the user instructions. In this work, we introduce a simple and effective method, which we name GUIDE, that mechanistically increases attention scores in instruction tokens. To support this operation, we present Influence, a novel metric that highlights how the user's instructions propagate through the transformer layers and impact the LLM output. Our results show that GUIDE improves the accuracy of following instructions 29.4 % to 60.4%, outperforming natural prompting alternatives and Supervised Fine-Tuning up to 1M tokens.
format Preprint
id arxiv_https___arxiv_org_abs_2409_19001
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Pay Attention to What Matters
Silva, Pedro Luiz
de Domenico, Antonio
Maatouk, Ali
Ayed, Fadhel
Computation and Language
Artificial Intelligence
Despite the remarkable success of Large Language Models (LLMs), they still exhibit a limited capability to align their outputs to the user instructions. In this work, we introduce a simple and effective method, which we name GUIDE, that mechanistically increases attention scores in instruction tokens. To support this operation, we present Influence, a novel metric that highlights how the user's instructions propagate through the transformer layers and impact the LLM output. Our results show that GUIDE improves the accuracy of following instructions 29.4 % to 60.4%, outperforming natural prompting alternatives and Supervised Fine-Tuning up to 1M tokens.
title Pay Attention to What Matters
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2409.19001