Exploring System 1 and 2 communication for latent reasoning in LLMs

Julian Coda-Forno, Zhuokai Zhao, Qiang Zhang, Dipesh Tamboli, WEIWEI LI, Xiangjun Fan, Lizhu Zhang, Eric Schulz, Hsiao-Ping Tseng

Keywords: LLMs, Latent reasoning, Cache augmentation, Latent communication

TL;DR: We explore dual-architecture latent reasoning—separating a Base from a Coprocessor (Liu et al. + two communication variants)—to see if it yields genuine reasoning; in our tests, current designs mostly buy extra compute, not robust reasoning.

Abstract:

Should LLM reasoning live in a separate coprocessor, or within a single model that uses the same forward pass and representational space? We study dual-architecture latent reasoning, where a fluent Base exchanges latent messages with a Coprocessor, and test two hypotheses aimed at improving latent communication over Liu et al. (2024b): (H1) increase channel capacity; (H2) learn communication via
joint finetuning. Under matched latent-token budgets on GPT-2 and Qwen-3, H2 is consistently strongest while H1 yields modest gains. A unified soft-embedding baseline—a single model with the same forward pass and shared representations, using the same latent-token budget—nearly matches H2 and surpasses H1, suggesting current dual designs mostly add compute rather than qualitatively improving reasoning. Across GSM8K, ProsQA, and a Countdown stress test with increasing branching factor, scaling the latent-token budget beyond small values fails to improve robustness. Latent analyses show overlapping subspaces with limited specialization, consistent with weak reasoning gains. We conclude dual-model latent reasoning remains promising in principle, but likely requires objectives and communication mechanisms that explicitly shape latent spaces for algorithmic planning.

Submission Number: 73