Task decomposition is a fundamental planning technique in artificial intelligence where a complex goal is broken down into a sequence of smaller, simpler, and executable steps. This allows an AI agent to formulate a coherent plan to solve problems that cannot be addressed in a single action. By creating a step-by-step plan, the agent can systematically work towards the final objective, often using tools or other sub-agents to complete each step. This method is central to advanced agent architectures for complex reasoning and problem-solving.

B. Static answer memorization involves retrieving pre-defined responses and is not a dynamic planning process for new goals.

C. Token truncation is a technical constraint for managing context window size, not a problem-solving strategy.

D. Simple keyword highlighting is a surface-level text processing feature that does not involve reasoning or planning.

1. Yao, S., et al. (2023). "ReAct: Synergizing Reasoning and Acting in Language Models." arXiv preprint arXiv:2210.03629. This influential paper introduces an agent architecture that explicitly decomposes tasks into thought/reasoning steps and action steps. (DOI: https://doi.org/10.48550/arXiv.2210.03629)

2. Russell, S. J., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed.). Chapter 3, "Solving Problems by Searching". The chapter introduces problem formulation, which involves breaking down a goal into a sequence of actions.

3. Stanford University, CS221: Artificial Intelligence: Principles and Techniques. The course syllabus covers "Search and Planning," which includes algorithms and methods for decomposing problems into state transitions and action sequences.