← Back to all news
arXiv Machine Learning By Leo Raphael Rodrigues, Jian Kang

Dissecting Hierarchical Reasoning Models: A Mechanistic Study

Read the original on arXiv Machine Learning →

The paper investigates Hierarchical Reasoning Models (HRM), a class of hierarchical Transformer-based latent reasoning models, across Sudoku, Maze, and ARC-AGI-2 tasks. By comparing HRM to Transformer baselines, applying causal interventions on recurrent states, and conducting linear probe and sparse autoencoder ablations, the authors find that recurrent models outperform one-pass baselines, that high- and low-level states contribute differently across tasks, and that ablations of sparse autoencoder features cause larger behavioral changes than probe-direction ablations. The study concludes that HRM implements constraint‑aware iterative refinement on a puzzle‑specific solution state, with component contributions varying without a compact, causally important feature set.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv Machine Learning
Aug 26

Steering Recurrent Reasoners at Inference Time with Readout Feedback

arXiv:2608.24136v1 Announce Type: new Abstract: Recurrent models, which repeatedly update latent states with shared computation blocks, have emerged as powerful architectures for solving complex reas...

By Shunsuke Kamiya, Masanori Koyama, Seongcheol Jeong, Fumiya Uchiyama, Kenji Kubo, Kohei Hayashi, Masahiro Suzuki, Yutaka Matsuo
More like this →
arXiv AI
Sep 2

Latent Recurrent Thoughts: Recurrent Refinement of Proposed Latents for Reasoning with Frozen LLMs

Latent Recurrent Thoughts (LRT) proposes a method for reasoning with frozen large language models by operating in the model’s continuous representation space. A small auxiliary network generates initial latent vectors, which a tiny recurrent reasoner refines over multiple steps, decoupling computational depth from model size. Experiments on symbolic and natural‑language reasoning tasks show that LRT outperforms prior frozen‑decoder continuous‑space methods and chain‑of‑thought prompting while using far less inference compute.

By Zhaoliang Chen, Jie Fu
llmsbenchmarks
More like this →
arXiv AI
Jul 29

Penelope: Localized Latent Recurrence for Efficient Structured Reasoning

arXiv:2607. 25915v1 Announce Type: new Abstract: Complex structured reasoning tasks often require additional computation, yet current language models obtain it mainly by increasing parameter scale or by serializing intermediate steps as chain-of-thought (CoT) tokens.

By Yutong Chen, Shouqian Shi, Xinran Liu, Haochen Wang, Jiaying Wang, Tianxing Xu, Yuanxi Wang, Zirui Ding
llmsdiffusionefficiencybenchmarks
More like this →
arXiv AI
Jul 31

Probing the Origins of Reasoning Performance: Representational Quality for Mathematical Problem-Solving in RL vs. SFT Fine-Tuned Models

arXiv:2607. 26119v1 Announce Type: new Abstract: Large reasoning models trained via reinforcement learning (RL) have been increasingly shown to outperform their supervised fine-tuned (SFT) counterparts on mathematical reasoning tasks; Yet the mechanistic basis for this advantage remains unclear.

By Antyabha Rahman, Akshaj Gurugubelli, Omar Ankit, Kevin Zhu, Aishwarya Balwani
reinforcement-learningfine-tuningbenchmarks
More like this →