A Survey of LLM Alignment: Instruction Understanding, Intention Reasoning, and Reliable Generation

TL;DR

This survey reviews the challenges faced by large language models in understanding instructions, reasoning about user intentions, and generating reliable, ethical responses, highlighting current solutions and future research directions.

Contribution

It provides a comprehensive classification and evaluation of existing benchmarks and solutions addressing LLM challenges in instruction understanding, intention reasoning, and reliable generation.

Findings

LLMs struggle with long contexts and multi-round instructions.

Inconsistent reasoning and understanding ambiguous commands are common issues.

Current benchmarks help evaluate and improve LLM performance in challenging scenarios.

Abstract

Large language models have demonstrated exceptional capabilities in understanding and generation. However, in real-world scenarios, users' natural language expressions are often inherently fuzzy, ambiguous, and uncertain, leading to challenges such as vagueness, polysemy, and contextual ambiguity. This paper focuses on three challenges in LLM-based text generation tasks: instruction understanding, intention reasoning, and reliable dialog generation. Regarding human complex instruction, LLMs have deficiencies in understanding long contexts and instructions in multi-round conversations. For intention reasoning, LLMs may have inconsistent command reasoning, difficulty reasoning about commands containing incorrect information, difficulty understanding user ambiguous language commands, and a weak understanding of user intention in commands. Besides, In terms of Reliable Dialog Generation, LLMs may have unstable generated content and unethical generation. To this end, we classify and analyze the performance of LLMs in challenging scenarios and conduct a comprehensive evaluation of existing solutions. Furthermore, we introduce benchmarks and categorize them based on the aforementioned three core challenges. Finally, we explore potential directions for future research to enhance the reliability and adaptability of LLMs in real-world applications.