Large language model agents struggle to pick and follow plans consistently
Do LLM Agents Execute the Plans They Declare? From Planning-Mode Declaration to Pattern-Specific Execution
Artificial IntelligenceMachine Learning
Summary
This paper looks at how language-based AI agents make plans and carry them out in tasks that take many steps. The authors find that these AI can choose a plan but often fail to follow it exactly, especially when the plans are long. They introduce a system that classifies planning into four types and matches each type with a specialized way to execute the plan, which helps the AI perform better. However, current AI still isn’t very good at picking the best plan type for each task, so there is room for improvement.
What this means in practice
- •For ai system engineers: Improve long-step task performance by using planning-mode classification and routing to ensure plans are executed as declared.
- •For robotics developers: Enhance robot task execution reliability by matching robot commands to planning patterns with specialized executors based on this research.
Authors
Subba Reddy Oota, Francisco Herrera, Jordi Cabot Sagrera, Marcos López de Prado, Shadab Khan
Abstract
Large language models (LLMs) enable agents to solve long-horizon tasks by generating a plan and then executing it in an environment. However, successful planning requires two distinct capabilities: selecting an appropriate plan for the task and executing it faithfully. Existing planner--executor systems can fail at either stage, while final task success alone cannot distinguish selection from execution failures. We therefore study the Plan Declaration--Execution Gap and introduce Planning-as-Routing, where an LLM declares one of four planning modes: Predefined, Sequential, Hierarchical, or Search, and a deterministic router dispatches the task to the corresponding pattern-specific executor. Across four benchmarks and three LLMs, we find three consistent patterns. First, generic Plan+ReAct often fails to preserve declared planning structure, especially for longer plans: across three benchmarks, only (22)--(45%) of trajectories preserve it, whereas pattern-specific executors enforce the intended structure. Second, planning-mode effectiveness varies across environments and models: Search performs best on ALFWorld, Hierarchical on SWE-bench, and the strongest pattern can vary across models within the same benchmark. Third, the largest gains come from execution: pattern-specific executors improve task success from (0.48) to (0.92) on ALFWorld and from (0.36) to (0.44) on SWE-bench Verified over Plan+ReAct. Current LLMs, however, do not reliably select the strongest mode for each task, although few-shot examples improve selection in some benchmark--model combinations. Overall, reliable agent planning requires both effective mode selection and faithful execution: routing substantially closes the execution gap, while task-specific mode selection remains open.