# When to Re-Plan: Subgoal Persistence in Hierarchical Latent Reasoning

arXiv: https://arxiv.org/abs/2606.03741 · HTML: https://mlsanity.com/paper/2606.03741/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 9 references that are in the map.
- First posted on arXiv 2 Jun 2026.

## Builds on (most cited first, 5 of 5)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Adaptive Computation Time for Recurrent Neural Networks](https://mlsanity.com/paper/1603.08983/) (2016)
- [FeUdal Networks for Hierarchical Reinforcement Learning](https://mlsanity.com/paper/1703.01161/) (2017)
- [Hierarchical Reasoning Model](https://mlsanity.com/paper/2506.21734/) (2025)
- [The ConceptARC Benchmark: Evaluating Understanding and Generalization in the ARC Domain](https://mlsanity.com/paper/2305.07141/) (2023)

## Built on by (most cited first, 1 of 1)

- [Loop the Loopies!](https://mlsanity.com/paper/2607.16051/) (2026)

## Abstract (arXiv)

Long-horizon reasoning requires a system to commit to medium-horizon intent without becoming rigid: re-plan too often and computation never coheres into multi-step structure; commit too long and the plan goes stale. We study this stability-adaptivity tradeoff in the latent reasoning setting, where multi-step computation occurs inside hidden state rather than externalized token traces. We extend the Hierarchical Reasoning Model (HRM) with a feudal-style manager-worker interface: a slow high-level module periodically emits a normalized directional subgoal that persists for P low-level steps, biasing the worker's hidden-state updates and supplying an intrinsic cosine alignment loss. On ARC and ConceptARC, we find that subgoal persistence -- not subgoal injection alone -- is the central knob: moderate periods P in [3, 6] consistently outperform both very frequent (P=1) and very long horizons, with a clear minimum LM loss at P=3 (1.544 vs. 1.674 at P=1, 1.640 baseline; replicated over 5 seeds at mean 1.595, std 0.045). The intrinsic alignment weight lambda shows a complementary narrow optimum (lambda approximately 0.05). A controlled ablation at past-sweet-spot lambda isolates learned directional structure -- not architectural capacity or auxiliary loss alone -- as the source of interference when the alignment signal exceeds its optimum. Together these findings implicate a design principle for compositional planning in latent reasoning systems: medium-horizon intent must be coherent across enough computational steps for compositional structure to form.