task decomposition
In plain wordsBreaking a big, vague job into smaller, concrete steps that can each be done and checked — the way a recipe splits a meal into stages.
The splitting of a complex goal into smaller subtasks, with their dependencies and order, that can be solved more reliably one at a time or in parallel. In language-model systems decomposition may be done by prompting (least-to-most or plan-and-solve prompting), by a planner model in plan-and-executeplan-and-execute, n.: An agent design that first writes out a complete step-by-step plan for a task, then carries out the steps one by one, revising the plan only when something goes wrong. agents, at run time by an orchestrator in the orchestrator–worker patternorchestrator–worker pattern, n.: A way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report., or by the humans who design an agentic workflowagentic workflow, n.: A multi-step process in which AI models draft, check, use tools and revise their work over several rounds, instead of answering in one shot; often the steps are laid out by developers in advance.. Good decompositions yield subtasks with clear inputs, outputs and success checks; poor ones lose context between pieces or produce parts that are individually easy but jointly inconsistent. Decomposition is also central to scalable oversightscalable oversight, n.: The problem of how people can check and correct AI systems whose work is too difficult, fast or plentiful for humans to judge reliably alone — usually by enlisting AI to help with the checking. proposals such as iterated amplificationiterated amplification, n.: A proposed training scheme in which a person, helped by many copies of an AI, solves harder problems than either could alone; a new AI then learns to imitate that team, and the cycle repeats., where a hard question is split into sub-questions that humans can evaluate.
The agent's task decomposition was fine; it failed because two subtasks edited the same config file.
Origin
A classical idea in problem solving and planning: Allen Newell and Herbert Simon's means–ends analysis in the General Problem SolverGeneral Problem Solver, prop. n.: A 1950s program meant to solve any well-defined puzzle by repeatedly comparing where it was with where it wanted to be and choosing steps that closed the gap. (late 1950s) and hierarchical task networkhierarchical task network, n.: A planning method that breaks a big task, such as 'travel to Paris', into smaller tasks, such as 'book a flight' and 'pack', using stored recipes, until only simple actions remain. planning from the 1970s. For language models, least-to-most prompting (Denny Zhou et al., 2022) and Plan-and-Solve prompting (Lei Wang et al., 2023) made explicit decomposition a prompting technique; Andrew Ng listed planning among four agentic design patterns in 2024.
Related
- plan-and-executeplan-and-execute, n.: An agent design that first writes out a complete step-by-step plan for a task, then carries out the steps one by one, revising the plan only when something goes wrong.
- orchestrator–worker patternorchestrator–worker pattern, n.: A way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report.
- hierarchical task networkhierarchical task network, n.: A planning method that breaks a big task, such as 'travel to Paris', into smaller tasks, such as 'book a flight' and 'pack', using stored recipes, until only simple actions remain.
- automated planningautomated planning, n.: The branch of AI that works out a sequence of actions to reach a goal, such as the steps a robot must take to assemble a part, from descriptions of what each action needs and what it changes.
- long-horizon tasklong-horizon task, n.: A job that takes many steps and a long time — hours or days of work rather than a single answer — so an AI must stay on track and recover from its own mistakes.
- chain of thoughtchain of thought, n.: Getting an AI to write out its reasoning step by step before giving an answer, the way a student shows their working. This often makes answers to math and logic problems more accurate.
Sources
- Wang, L., Xu, W., Lan, Y., Hu, Z., Lan, Y., Lee, R. K.-W. & Lim, E.-P. (2023). Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models. arXiv:2305.04091. arxiv.org/abs/2305.04091
- Zhou, D. et al. (2022). Least-to-Most Prompting Enables Complex Reasoning in Large Language Models. arXiv:2205.10625 (ICLR 2023).
Mentioned in
agentic workflowagentic workflow, n.: A multi-step process in which AI models draft, check, use tools and revise their work over several rounds, instead of answering in one shot; often the steps are laid out by developers in advance., long-horizon tasklong-horizon task, n.: A job that takes many steps and a long time — hours or days of work rather than a single answer — so an AI must stay on track and recover from its own mistakes., orchestrator–worker patternorchestrator–worker pattern, n.: A way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report., plan-and-executeplan-and-execute, n.: An agent design that first writes out a complete step-by-step plan for a task, then carries out the steps one by one, revising the plan only when something goes wrong., prompt chainingprompt chaining, n.: A way of getting a big job done with a relay of AI requests: each handles one stage and passes its result to the next — pull out the facts, write, polish — with automatic checks between stages., BabyAGIBabyAGI, prop. n.: A tiny 2023 program that kept an AI working through an ever-changing to-do list — doing a task, adding new ones, reprioritizing — and helped start the craze for autonomous AI agents., factored cognitionfactored cognition, n.: The idea that a hard thinking task can be broken into many small, separately checkable pieces, each simple enough for a person or an AI to handle in a short time., hierarchical reinforcement learninghierarchical reinforcement learning, n.: Reinforcement learning organized in layers, where a high-level part chooses goals or skills — 'go to the kitchen' — and lower-level parts handle the detailed actions to achieve them., least-to-most promptingleast-to-most prompting, n.: A prompting method that has an AI list the smaller questions a hard problem depends on, answer the easiest first, and feed each answer into the next, building up to the full solution.