Jet Expansions of Residual Computation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Chen, Yihong, Xu, Xiangxiang, Lu, Yao, Stenetorp, Pontus, Franceschi, Luca
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914968014684160
author Chen, Yihong
Xu, Xiangxiang
Lu, Yao
Stenetorp, Pontus
Franceschi, Luca
author_facet Chen, Yihong
Xu, Xiangxiang
Lu, Yao
Stenetorp, Pontus
Franceschi, Luca
contents We introduce a framework for expanding residual computational graphs using jets, operators that generalize truncated Taylor series. Our method provides a systematic approach to disentangle contributions of different computational paths to model predictions. In contrast to existing techniques such as distillation, probing, or early decoding, our expansions rely solely on the model itself and requires no data, training, or sampling from the model. We demonstrate how our framework grounds and subsumes logit lens, reveals a (super-)exponential path structure in the recursive residual depth and opens up several applications. These include sketching a transformer large language model with $n$-gram statistics extracted from its computations, and indexing the models' levels of toxicity knowledge. Our approach enables data-free analysis of residual computation for model interpretability, development, and evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2410_06024
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Jet Expansions of Residual Computation
Chen, Yihong
Xu, Xiangxiang
Lu, Yao
Stenetorp, Pontus
Franceschi, Luca
Machine Learning
Artificial Intelligence
Computation and Language
Symbolic Computation
We introduce a framework for expanding residual computational graphs using jets, operators that generalize truncated Taylor series. Our method provides a systematic approach to disentangle contributions of different computational paths to model predictions. In contrast to existing techniques such as distillation, probing, or early decoding, our expansions rely solely on the model itself and requires no data, training, or sampling from the model. We demonstrate how our framework grounds and subsumes logit lens, reveals a (super-)exponential path structure in the recursive residual depth and opens up several applications. These include sketching a transformer large language model with $n$-gram statistics extracted from its computations, and indexing the models' levels of toxicity knowledge. Our approach enables data-free analysis of residual computation for model interpretability, development, and evaluation.
title Jet Expansions of Residual Computation
topic Machine Learning
Artificial Intelligence
Computation and Language
Symbolic Computation
url https://arxiv.org/abs/2410.06024