Saved in:
Bibliographic Details
Main Author: Frydenlund, Arvid
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.13779
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910954255548416
author Frydenlund, Arvid
author_facet Frydenlund, Arvid
contents The recently introduced path-star task is a minimal task designed to exemplify limitations to the abilities of language models (Bachmann and Nagarajan, 2024). It involves a path-star graph where multiple arms radiate from a single starting node and each node is unique. Given the start node and a specified target node that ends an arm, the task is to generate the arm containing that target node. This is straightforward for a human but surprisingly difficult for language models, which did not outperform the random baseline. The authors hypothesized this is due to a deficiency in teacher-forcing and the next-token prediction paradigm. We demonstrate the task is learnable using teacher-forcing in alternative settings and that the issue is partially due to representation. We introduce a regularization method using structured samples of the same graph but with differing target nodes, improving results across a variety of model types. We provide RASP proofs showing the task is theoretically solvable. Finally, we find settings where an encoder-only model can consistently solve the task.
format Preprint
id arxiv_https___arxiv_org_abs_2410_13779
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Mystery of the Pathological Path-star Task for Language Models
Frydenlund, Arvid
Computation and Language
Machine Learning
The recently introduced path-star task is a minimal task designed to exemplify limitations to the abilities of language models (Bachmann and Nagarajan, 2024). It involves a path-star graph where multiple arms radiate from a single starting node and each node is unique. Given the start node and a specified target node that ends an arm, the task is to generate the arm containing that target node. This is straightforward for a human but surprisingly difficult for language models, which did not outperform the random baseline. The authors hypothesized this is due to a deficiency in teacher-forcing and the next-token prediction paradigm. We demonstrate the task is learnable using teacher-forcing in alternative settings and that the issue is partially due to representation. We introduce a regularization method using structured samples of the same graph but with differing target nodes, improving results across a variety of model types. We provide RASP proofs showing the task is theoretically solvable. Finally, we find settings where an encoder-only model can consistently solve the task.
title The Mystery of the Pathological Path-star Task for Language Models
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2410.13779