Search papers, labs, and topics across Lattice.
The paper introduces Recoverable Program-of-Thought (RePoT), a method that uses verified replay and checkpointing to recover from invalid actions in Program-of-Thought (PoT) trajectories by resuming execution from the last valid state. RePoT significantly improves PoT's success rate by up to 11 percentage points across various closed and open-source models on planning benchmarks like PuzzleZoo and PlanBench, while incurring minimal overhead (at most one extra LLM call on ~14% of problems). Analysis on a controlled recovery benchmark (Derail-550) reveals that checkpoint information, rather than the specific verified-prefix tail, is the key signal for successful recovery.
LLMs can recover from planning failures with just one extra call, achieving near-perfect accuracy by replaying and repairing their "train of thought" from verified checkpoints.
One-shot Program-of-Thought (PoT) emits a Python program that prints a primitive-action plan; a single invalid action silently invalidates the trajectory. We introduce RePoT (Recoverable PoT): a deterministic verified replay that walks the plan through the environment to its first invalid transition, then one LLM call that resumes from the verified prefix. RePoT costs at most one extra LLM call on the ~14% of problems where PoT fails. RePoT beats PoT by +3 to +11pp across four closed-model configurations on PuzzleZoo-775 and peaks at 96.9% vs 86.3% on gpt-5.4-mini-medium; against the matched-budget PoT-retry baseline, RePoT wins decisively on Gemini (+3.8pp, 95% CI [+2.2,+5.4]), is within sampling noise on GPT-medium and Claude, and loses on GPT-mini -- a capability-scaling pattern we begin to address with Adaptive RePoT, a rule-based dispatcher that routes between suffix repair and a fresh PoT retry based on verified-prefix length (preliminary). We replicate on PlanBench Blocksworld (+1.1 to +11.4pp) and on four open-weights models (+3.3 to +20.0pp on three of four). On Derail-550, our controlled recovery benchmark, every condition with access to checkpoint information clears>=30% on GPT-medium and>=70% on Gemini, vs<=3.1% for error-only feedback -- showing that checkpoint information, not the specific verified-prefix tail, is the load-bearing recovery signal.
