SIRCL: Stabilizing Implicit Reasoning via Centripetal Latent Trajectories

Jan 1, 2026·
Yuwen Hao
Yuwen Hao
Equal contribution
,
Jindong Li
,
Changjing Zhu
,
Menglin Yang
· 1 min read
Image credit: Unsplash
Abstract
This work studies latent-space instability and semantic drift in implicit chain-of-thought reasoning. It models latent tokens as continuous reasoning trajectories and proposes a lightweight geometric regularizer that constrains excessive trajectory deviation during training.
Type
Publication
Under Review
Status
Open access
publications
Note

This paper is currently under review.

SIRCL investigates instability in implicit chain-of-thought reasoning, where latent-token trajectories may drift away from the problem core as reasoning steps increase. The method introduces a sample-level center and hinge-style trust-region loss to regularize excessive latent displacement during training while keeping inference cost unchanged.

Yuwen Hao
Authors
MPhil Student in Artificial Intelligence
I am an MPhil student in Artificial Intelligence at the Hong Kong University of Science and Technology (Guangzhou), working with Assistant Professor Menglin Yang in DIGAI Lab. My current research focuses on continuous latent-space reasoning, implicit chain-of-thought reasoning, and language model reasoning systems.