An Incomplete Loop: Instruction Inference, Instruction Following, and In-context Learning in Language Models

TL;DR

This paper investigates how language models perform instruction inference, instruction following, and in-context learning, revealing their distinct reasoning capabilities and limitations across different tasks and models.

Contribution

It provides a comparative analysis of different reasoning types in LMs, highlighting the non-systematic nature of their learning mechanisms and dissociations between inference and learning.

Findings

LMs can learn from few-shot prompts without explaining their rules

They can infer task descriptions but fail to learn from human descriptions

Different reasoning capabilities are invoked by similar prompting procedures

Abstract

Modern language models (LMs) can learn to perform new tasks in different ways: in instruction following, the target task is described explicitly in natural language; in few-shot prompting, the task is specified implicitly with a small number of examples; in instruction inference, LMs are presented with in-context examples and are then prompted to generate a natural language task description before making predictions. Each of these procedures may be thought of as invoking a different form of reasoning: instruction following involves deductive reasoning, few-shot prompting involves inductive reasoning, and instruction inference involves abductive reasoning. How do these different capabilities relate? Across four LMs (from the gpt and llama families) and two learning problems (involving arithmetic functions and machine translation) we find a strong dissociation between the different types of reasoning: LMs can sometimes learn effectively from few-shot prompts even when they are unable to explain their own prediction rules; conversely, they sometimes infer useful task descriptions while completely failing to learn from human-generated descriptions of the same task. Our results highlight the non-systematic nature of reasoning even in some of today's largest LMs, and underscore the fact that very different learning mechanisms may be invoked by seemingly similar prompting procedures.