llm-guided-narrow-space-parking-where-optimal-control-solver-supplies-pass-fail CLM-0424For LLM-guided narrow-space parking where an optimal-control solver supplies pass/fail evidence, growing a decision knowledge base offline from accumulated solver and validation failures — retaining an entry only when re-planning then yields a feasible trajectory — raises planning success from 85.1% to 94.7% and closed-loop success from 74.6% to 84.2% over an empty knowledge base, with model weights frozen.
Capability: Improving its own scaffolding from failures
Observed on
38 simulated confined parking scenarios, car-like kinematic model, all LLM components gpt-5.6-sol, bounded online replanning budget of three attempts; one qualitative differential-.
Sources
- Validated by: Ablation on 38 scenarios against the same pipeline with an empty knowledge base: planning success 94.7% vs 85.1%, closed-loop success 84.2% vs 74.6%; also compared against HA*, Direct OCP and HA*+OCP baselines.. Measured: single ablation table on 38 simulated scenarios with one LLM, one vehicle model, one controller. Isolates the knowledge base against an empty base with other settings fixed, but no repeated trials, no variance reported, and no test of whether hand-written entries would do as well. Real-robot evidence is a single qualitative run.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- When training an LLM with policy-gradient RL on tasks that decompose into many dependent sub-problems, rewarding each trajectory segment by how many reference reasoning points it newly reaches beats a sparse final-answer reward by a margin that widens with task horizon, and still learns on problem sets where the base policy never samples a success and sparse reward gives no gradient at all.Completing tasks that take many steps or many hours · reviewed by AI
- For long-horizon embodied task planning where a symbolic model defines action preconditions and effects, correcting a violated plan by edits the action model itself implies (inserting the missing navigate or open) raises plan success more per LLM call than returning the violation to the LLM for replanning, and the gain persists for frontier models as well as compact ones.Planning a sequence of actions before acting · reviewed by AI
- For open-source LLM agents driving a ROS robot, having the model emit a pseudo-code tool-call plan up front, storing it in a persistent scratchpad, and then restricting each tool invocation to the plan's next step raises task success most on the longest multi-step tasks and for the smaller open models (GPT-OSS-20B on the hardest task rose from 35% with a persistent plan alone to 55% with the constraint), while a strong proprietary model gains little.Completing tasks that take many steps or many hours · reviewed by AI
- For training-free MLLM-driven navigation (instruction-following and object-goal) in simulated indoor environments, gating each planner-proposed action on evidential support, geometric reachability and subgoal consistency before dispatch — and refusing a stop request unless semantic, geometric and progress checks all pass — raises success rate over letting the planner act directly, but the stop check buys completion at some cost in path efficiency.Turning plans into physical action · reviewed by AI
- On 102 IPC Blocksworld instances with a 180-second budget, a classical planner (Fast Downward) solved 85%, a model planning directly solved 64%, and the same model choosing one action at a time against a PDDL simulator with observe, reset and retry solved 67%, a consistent but small three-point gain at about 5.7 times the token cost, so interactive simulation helps a little and classical search still wins where a formal model exists.Planning a sequence of actions before acting · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: low. Falsifier as drafted: An ablation on the same scenarios showing the evolved knowledge base gives no planning or closed-loop gain over an empty base once attempt budget and compute are matched, or that the gain is attributable to replanning attempts rather than the retained entries.