Research

CosmicFish-HRM

A study of adaptive reasoning in compact language models. A hierarchical recurrent mechanism lets the model decide how much computation each input deserves, continuing or halting its reasoning rather than spending a fixed amount on every request.

The model

82.77M parameters · transformer · GQA · RoPE · SwiGLU · RMSNorm

Adaptive depth with a learned halt/continue mechanism, up to 16 reasoning steps.

Adaptive computation

Reported mean reasoning steps by task:

Simple factual completion2.78
Scientific concepts4.77
Logical reasoning7.03
Multi-step arithmetic8.40
Paper

CosmicFish-HRM: Adaptive Reasoning via Hierarchical Recurrent Mechanisms in Compact Language Models