# What Makes Instruction Learning Hard? An Investigation and a New Challenge in a Synthetic Environment

arXiv: https://arxiv.org/abs/2204.09148 · HTML: https://mlsanity.com/paper/2204.09148/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 31 references that are in the map.
- First posted on arXiv 19 Apr 2022.

## Builds on (most cited first, 23 of 23)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [Multitask Prompted Training Enables Zero-Shot Task Generalization](https://mlsanity.com/paper/2110.08207/) (2021)
- [Aligning AI With Shared Human Values](https://mlsanity.com/paper/2008.02275/) (2020)
- [Cross-Task Generalization via Natural Language Crowdsourcing Instructions](https://mlsanity.com/paper/2104.08773/) (2021)
- [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://mlsanity.com/paper/2105.13626/) (2021)
- [Theoretical Limitations of Self-Attention in Neural Sequence Models](https://mlsanity.com/paper/1906.06755/) (2019)
- [Measuring Compositional Generalization: A Comprehensive Method on Realistic Data](https://mlsanity.com/paper/1912.09713/) (2019)
- [Transformers as Soft Reasoners over Language](https://mlsanity.com/paper/2002.05867/) (2020)
- [Compositional Generalization and Natural Language Variation: Can a Semantic Parsing Approach Handle Both?](https://mlsanity.com/paper/2010.12725/) (2020)
- [Adapting Language Models for Zero-shot Learning by Meta-tuning on Dataset and Prompt Collections](https://mlsanity.com/paper/2104.04670/) (2021)
- [The Turking Test: Can Language Models Understand Instructions?](https://mlsanity.com/paper/2010.11982/) (2020)
- [Learning from Task Descriptions](https://mlsanity.com/paper/2011.08115/) (2020)
- [What Do Neural Networks Learn When Trained With Random Labels?](https://mlsanity.com/paper/2006.10455/) (2020)
- [Does Pretraining for Summarization Require Knowledge Transfer?](https://mlsanity.com/paper/2109.04953/) (2021)
- [Unobserved Local Structures Make Compositional Generalization Hard](https://mlsanity.com/paper/2201.05899/) (2022)
- [Pushing the Limits of Rule Reasoning in Transformers through Natural Language Satisfiability](https://mlsanity.com/paper/2112.09054/) (2021)
- [Synthetic Datasets for Neural Program Synthesis](https://mlsanity.com/paper/1912.12345/) (2019)
- [On the Ability and Limitations of Transformers to Recognize Formal Languages](https://mlsanity.com/paper/2009.11264/) (2020)
- [Formal Language Theory Meets Modern NLP](https://mlsanity.com/paper/2102.10094/) (2021)
- [Dyna-bAbI: unlocking bAbI's potential with dynamic synthetic benchmarking](https://mlsanity.com/paper/2112.00086/) (2021)
- [Understanding Robust Generalization in Learning Regular Languages](https://mlsanity.com/paper/2202.09717/) (2022)

## Built on by (most cited first, 9 of 9)

- [The Parallelism Tradeoff: Limitations of Log-Precision Transformers](https://mlsanity.com/paper/2207.00729/) (2022)
- [Improving Instruction-Following in Language Models through Activation Steering](https://mlsanity.com/paper/2410.12877/) (2024)
- [In-Context Language Learning: Architectures and Algorithms](https://mlsanity.com/paper/2401.12973/) (2024)
- [Simplicity Bias in Transformers and their Ability to Learn Sparse Boolean Functions](https://mlsanity.com/paper/2211.12316/) (2022)
- [Active Instruction Tuning: Improving Cross-Task Generalization by Training on Prompt Sensitive Tasks](https://mlsanity.com/paper/2311.00288/) (2023)
- [Do Models Really Learn to Follow Instructions? An Empirical Study of Instruction Tuning](https://mlsanity.com/paper/2305.11383/) (2023)
- [Learning Instructions with Unlabeled Data for Zero-Shot Cross-Task Generalization](https://mlsanity.com/paper/2210.09175/) (2022)
- [Induced Natural Language Rationales and Interleaved Markup Tokens Enable Extrapolation in Large Language Models](https://mlsanity.com/paper/2208.11445/) (2022)
- [SIP: Injecting a Structural Inductive Bias into a Seq2Seq Model by Simulation](https://mlsanity.com/paper/2310.00796/) (2023)

## Abstract (arXiv)

The instruction learning paradigm -- where a model learns to perform new tasks from task descriptions alone -- has become popular in general-purpose model research. The capabilities of large transformer models as instruction learners, however, remain poorly understood. We use a controlled synthetic environment to characterize such capabilities. Specifically, we use the task of deciding whether a given string matches a regular expression (viewed as an instruction) to identify properties of tasks, instructions, and instances that make instruction learning challenging. For instance, we find that our model, a fine-tuned T5-based text2text transformer, struggles with large regular languages, suggesting that less precise instructions are challenging for models. Additionally, instruction executions that require tracking longer contexts of prior steps are also more difficult. We use our findings to systematically construct a challenging instruction learning dataset, which we call Hard RegSet. Fine-tuning on Hard RegSet, our large transformer learns to correctly interpret only 65.6% of test instructions (with at least 90% accuracy), and 11%-24% of the instructions in out-of-distribution generalization settings. We propose Hard RegSet as a challenging instruction learning task, and a controlled environment for studying instruction learning.