The Plan Declaration--Execution Gap is studied and Planning-as-Routing is introduced, 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.
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.
Large language models (LLMs) perform well on many stepwise reasoning tasks, yet their ability to construct reliable plans over long horizons remains difficult to measure. Existing evaluations often use compact symbolic domains or allow execution-time feedback, making it harder to isolate failures in end-to-end plan con...
P. Anokhin, Roman Khalikov, Stefan Rebrikov et al.· IEEE Access· 0 citations
Planning at run time, as in robots that re-plan while they act, needs a planner that reads the PDDL its users write (durative actions, numeric resources, time windows, continuous change), answers quickly, and delivers only validated plans. Specialised planners are strong on individual fragments of PDDL, but they differ...
F. Rico, Esther Aguado González, Francisco Miguel Moreno Olivo et al.· 0 citations
Capability-Grounded Agentic Planning (CGAP) is presented, which compiles the tool catalog deterministically into a PDDL model, uses an LLM to turn the request into goals, and solves the model with a classical planner.
Shikha Shrestha, William Yeoh, Ning Zhang et al.· 0 citations
LLM agents that execute multi-step tool calls must satisfy two objectives simultaneously: completing the user's task (utility) and conforming to domain policies and correctness constraints (safety). These objectives are in tension -- blocking an unsafe action prevents a violation but can leave the agent stranded with n...
Mike He, Ankush Desai, Sharad Malik et al.· 0 citations
The proposed Plan-Then-Verify (PTV), a verifier-first control pattern in which a candidate workflow is checked against a documented knowledge base before execution begins, significantly improves static SOP conformance on several model families, while remaining close to Plan-and-Execute in token cost.
Emil Benedykciuk, Marcin Denkowski, Marek Jezior et al.· Applied Sciences· 0 citations
The challenges of reliable asynchronous task execution in LLM agents are highlighted and the need to evaluate how agents handle incoming results alongside end-to-end task success is highlighted.
De-Tao Zeng, Gui-Lin Ou, Nuo Chen et al.· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 13, 2026
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
MIT News · Artificial Intelligence· news.mit.eduSep 29, 2026
Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.