按年被提到
20212023 · 2 次2026
大模型与语言模型›模型表达力与计算理论·arXiv:2102.13019 · cs.CL

Investigating the Limitations of Transformers with Simple Arithmetic Tasks

Rodrigo Nogueira, Zhiying Jiang, Jimmy Lin · 2021
开始阅读 ↗PDFalphaXiv 讨论
被提及 · 2 人首次 2023-02-06 · 最近 2023-03-30
2不同的人
2活跃月份
2讨论串
57被引(OpenAlex)
arXiv 摘要(候选层不写导读)

The ability to perform arithmetic tasks is a remarkable trait of human intelligence and might form a critical component of more complex reasoning tasks. In this work, we investigate if the surface form of a number has any influence on how sequence-to-sequence language models learn simple arithmetic tasks such as addition and subtraction across a wide range of values. We find that how a number is represented in its surface form has a strong influence on the model's accuracy. In particular, the model fails to learn addition of five-digit numbers when using subwords (e.g., "32"), and it struggles to learn with character-level representations (e.g., "3 2"). By introducing position tokens (e.g., "3 10e1 2"), the model learns to accurately add and subtract numbers up to 60 digits. We conclude that modern pretrained language models can easily learn arithmetic from very few examples, as long as we use the proper surface representation. This result bolsters evidence that subword tokenizers and positional encodings are components in current transformer designs that might need improvement. Moreover, we show that regardless of the number of parameters and training examples, models cannot learn addition rules that are independent of the length of the numbers seen during training. Code to reproduce our experiments is available at https://github.com/castorini/transformers-arithmetic

候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
评论versteegen · 2023-02

“I wrote that double descent is widespread, not grokking. Of course a transformer can't do multiplication or any other kind of operation on an infinite set of numbers, because it has only bounded depth which limits the number of steps it can emulate of any algorithm. But I think I see how I could build a transformer by hand that could multiply any two 4-d…”

Understanding the limits of large language models原评论 ↗
评论spyder · 2023-03

“That's also an interesting one, but I think I found it, but it's a little different from what I remembered because it's not just the scientific notation that helps but more so the addition of positional tokens: https://arxiv.org/abs/2102.13019 The essence of it from the abstract: ". In particular, the model fails to learn addition of five-digit numbers…”

Cerebras-GPT vs. LLaMA AI Model Performance Comparison原评论 ↗