AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Mixture of Cognitive Reasoners: Modular Reasoning with Brain-Like Specialization

Badr AlKhamissi, C. Nicolò De Sabbata, Greta Tuckute, Zeming Chen, Martin Schrimpf, Antoine Bosselut

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Reasoning mechanisms and failure modes · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add brain-inspired Stage-1/2 specialization curriculum (MiCRo) over unspecialized mixture-of-blocks (ablation)MoB · GSM8Kaccuracy (%): 30.5 -> 34.7 (Llama-1B); 51.6 -> 57.2 (Llama-3B)Improved
MiCRo versus dense model on reasoning benchmarksLlama-3.2-1B/3B · GSM8K, MMLU, BBH, Minervaaccuracy (%): GSM8K 36.8 -> 34.7 (1B); MMLU 29.7 -> 31.2 (1B); MMLU 48.6 -> 45.4 (3B)Mixed
ablate the Logic expert (ablation)MiCRo · MATH / GSM8KaccuracyHurt
ablate the Social expert (ablation)MiCRo · math benchmarksaccuracyImproved
ablate the Language expert (ablation)MiCRo · all benchmarksaccuracyHurt
apply DPO after MiCRo training versus denseLlama-3.2-1B · DPO 4-benchmark averageaverage accuracy (%): 25.4 -> 26.8Improved
MiCRo with medical instruction tuning versus denseLlama-3.2-1B · Medical averageaverage accuracy (%): 41.9 -> 42.2Improved
use MoE (FFN-only experts) instead of mixture-of-blocks (ablation)MiCRo · specialization emergencerouting specializationHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy