TRM-Planner: Offline Target Planning and Distillation for Tiny Recursive Models

Tiny Recursive Models (TRMs) perform iterative reasoning with an Adaptive Computation Time (ACT)-style loop, but their effective supervision can be target-sensitive to refinement depth and halting calibration, and their halting behavior can be difficult to tune. We introduce TRM-Planner, a two-stage teachercache distillation recipe that shifts compute to an offline teacher-cache stage. A frozen TRM checkpoint is unrolled for multiple refinement steps and stochastic rollouts; for each instance we cache a small set of teacher entries (tokens, logits, step index, and quality metadata). A student TRM is then trained with the standard TRM objective plus a distillation loss computed from cached entries. Across Sudoku-Extreme and ARC-AGI-1/2, TRM-Planner shows an improvement over our reproduced TRM baseline while leaving studenttime inference unchanged. On ARC1/ARC2 with 7M parameters, the two-attempt accuracy (pass@2) increases from 43.1%→48.1% and 6.7%→9.2%, respectively.