In-context learning and chain-of-thought prompting have demonstrated surprising performance improvements on mathematical reasoning benchmarks. Therefore, understanding the underlying factors enabling these capabilities is crucial. However, the specific aspects of pretraining data that equip models with mathematical reasoning capabilities remain largely unexplored and are less studied systematically. In this study, we identify subsets of model pretraining data that contribute to math reasoning ability of the model, and evaluate it on several mathematical operations (e.g. addition, multiplication) and tasks (e.g. the asdiv dataset). We measure the importance of such subsets by continual training of the model on pretraining data subsets, and then we quantify the change in performance on the mathematical benchmark to assess their importance. If a subset results in an improved performance, we conjecture that such subset contributes to a model's overall mathematical ability. Our results unveil that while training on math-only data contributes to simple arithmetic abilities, it does not solely explain performance on more complex reasoning abilities like chain-of-thought reasoning. We also find that code data contributes to chain-of-thought reasoning while reducing the arithmetic performance.
more »
« less
Teaching Arithmetic to Small Transformers
Large language models like GPT-4 exhibit emergent capabilities across general-purpose tasks, such as basic arithmetic, when trained on extensive text data, even though these tasks are not explicitly encoded by the unsupervised, next-token prediction objective. This study investigates how even small transformers, trained from random initialization, can efficiently learn arithmetic operations such as addition, multiplication, and elementary functions like square root, using the next-token prediction objective. We first demonstrate that conventional training data is not the most effective for arithmetic learning, and simple formatting changes can significantly improve accuracy. This leads to sharp phase transitions as a function of training data scale, which, in some cases, can be explained through connections to low-rank matrix completion. Building on prior work, we then train on chain-of-thought style data that includes intermediate step results. Even in the complete absence of pretraining, this approach significantly and simultaneously improves accuracy, sample complexity, and convergence speed. We also study the interplay between arithmetic and text data during training and examine the effects of few-shot prompting, pretraining, and parameter scaling. Additionally, we discuss the challenges associated with length generalization. Our work highlights the importance of high-quality, instructive data that considers the particular characteristics of the next-word prediction loss for rapidly eliciting arithmetic capabilities.
more »
« less
- Award ID(s):
- 2144994
- NSF-PAR ID:
- 10511455
- Publisher / Repository:
- ICLR 2024
- Date Published:
- Journal Name:
- International Conference on Learning Representations
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)Existing work on tabular representation-learning jointly models tables and associated text using self-supervised objective functions derived from pretrained language models such as BERT. While this joint pretraining improves tasks involving paired tables and text (e.g., answering questions about tables), we show that it underperforms on tasks that operate over tables without any associated text (e.g., populating missing cells). We devise a simple pretraining objective (corrupt cell detection) that learns exclusively from tabular data and reaches the state-of-the-art on a suite of table-based prediction tasks. Unlike competing approaches, our model (TABBIE) provides embeddings of all table substructures (cells, rows, and columns), and it also requires far less compute to train. A qualitative analysis of our model’s learned cell, column, and row representations shows that it understands complex table semantics and numerical trends.more » « less
-
A real-world text corpus sometimes comprises not only text documents, but also semantic links between them (e.g., academic papers in a bibliographic network are linked by citations and co-authorships). Text documents and semantic connections form a text-rich network, which empowers a wide range of downstream tasks such as classification and retrieval. However, pretraining methods for such structures are still lacking, making it difficult to build one generic model that can be adapted to various tasks on text-rich networks. Current pretraining objectives, such as masked language modeling, purely model texts and do not take inter-document structure information into consideration. To this end, we propose our PretrAining on TexT-Rich NetwOrk framework PATTON. PATTON1 includes two pretraining strategies: network-contextualized masked language modeling and masked node prediction, to capture the inherent dependency between textual attributes and network structure. We conduct experiments on four downstream tasks in five datasets from both academic and e-commerce domains, where PATTON outperforms baselines significantly and consistently.more » « less
-
Benjamin, Paaßen ; Carrie, Demmans Epp (Ed.)One of the areas where Large Language Models (LLMs) show promise is for automated qualitative coding, typically framed as a text classification task in natural language processing (NLP). Their demonstrated ability to leverage in-context learning to operate well even in data-scarce settings poses the question of whether collecting and annotating large-scale data for training qualitative coding models is still beneficial. In this paper, we empirically investigate the performance of LLMs designed for use in prompting-based in-context learning settings, and draw a comparison to models that have been trained using the traditional pretraining--finetuning paradigm with task-specific annotated data, specifically for tasks involving qualitative coding of classroom dialog. Compared to other domains where NLP studies are typically situated, classroom dialog is much more natural and therefore messier. Moreover, tasks in this domain are nuanced and theoretically grounded and require a deep understanding of the conversational context. We provide a comprehensive evaluation across five datasets, including tasks such as talkmove prediction and collaborative problem solving skill identification. Our findings show that task-specific finetuning strongly outperforms in-context learning, showing the continuing need for high-quality annotated training datasets.more » « less
-
Train Big, Then Compress: Rethinking Model Size for Efficient Training and Inference of TransformersIII, Hal Daumé (Ed.)Since hardware resources are limited, the objective of training deep learning models is typically to maximize accuracy subject to the time and memory constraints of training and inference. We study the impact of model size in this setting, focusing on Transformer models for NLP tasks that are limited by compute: self-supervised pretraining and high-resource machine translation. We first show that even though smaller Transformer models execute faster per iteration, wider and deeper models converge in significantly fewer steps. Moreover, this acceleration in convergence typically outpaces the additional computational overhead of using larger models. Therefore, the most compute-efficient training strategy is to counterintuitively train extremely large models but stop after a small number of iterations. This leads to an apparent trade-off between the training efficiency of large Transformer models and the inference efficiency of small Transformer models. However, we show that large models are more robust to compression techniques such as quantization and pruning than small models. Consequently, one can get the best of both worlds: heavily compressed, large models achieve higher accuracy than lightly compressed, small models.more » « less
