Article

Think Wider: Mitigating Latent Rank Collapse in Implicit Chain-of-Thought Reasoning featured image

Think Wider: Mitigating Latent Rank Collapse in Implicit Chain-of-Thought Reasoning

A spectral regularization method for improving latent-token diversity and mitigating rank collapse in implicit reasoning.

avatar
Yuwen Hao
•
SynWeaver: Website-Prior Task and Trajectory Co-Synthesis for Web Agents featured image

SynWeaver: Website-Prior Task and Trajectory Co-Synthesis for Web Agents

Website-prior task and trajectory co-synthesis for web agents.

ruitao-wang
•
SIRCL: Stabilizing Implicit Reasoning via Centripetal Latent Trajectories featured image

SIRCL: Stabilizing Implicit Reasoning via Centripetal Latent Trajectories

A lightweight geometric regularization method for stabilizing implicit reasoning trajectories without adding inference-time overhead.

avatar
Yuwen Hao
•

Learning Less Is More: Premature Upper-Layer Attention Specialization Hurts Language Model Pretraining

A study of premature upper-layer attention specialization during language model pretraining.

jinchang-zhu
•