AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

MAGO: Beyond Fixed Hyperparameters with Multi-Objective Pareto Optimization for Hybrid LLM Reasoning

Hongcheng Ding, Xuanze Zhao, Ruiting Deng, Shamsul Nahar Abdullah, Deshinta Arrova Dewi, LIU QINGYU

ICLR 2026 · Large language models › Reasoning and chain of thought · approach: Efficient and adaptive thinking

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes MAGO, a multi-objective Pareto optimization approach for hybrid LLM reasoning that decides adaptively between direct answering and chain-of-thought reasoning instead of using fixed hyperparameters.

Problem

Uniformly applying elaborate chain-of-thought reasoning to all queries wastes computation, since many problems can be solved directly; existing hybrid reasoning methods rely on static fixed hyperparameters for switching between reasoning modes.

Approach

Frames hybrid reasoning mode selection as a multi-objective problem solved via Pareto optimization, replacing fixed hyperparameters with a trade-off between accuracy and computational cost.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace fixed weight alpha with Pareto adaptive multi-objective weighting (MAGO)DeGRPO hybrid reasoning, 1.5B · AIME 2024Pass@1: 0.2506 -> 0.2741Improved
replace fixed weight alpha with MAGODeGRPO, 1.5B · MATH-500Pass@1: 0.8037 -> 0.8247Improved
replace fixed weight alpha with MAGODeGRPO, 1.5B · Minerva AlgebraPass@1: 0.9216 -> 0.9483Improved
use MAGO hybrid model instead of full long-CoT base modelDeepSeek-R1-1.5B · AIME 2024tokens per query: 18063 -> 7164Improved
use MAGO hybrid model instead of full long-CoT base modelDeepSeek-R1-1.5B · MATH-500Pass@1: 0.8608 -> 0.8247Hurt
use MAGO instead of QMath-1.5B on GSM8KQwen-1.5B · GSM8KPass@1: 0.8572 -> 0.8469No clear effect
apply MAGO trained on math to CommonsenseQADeGRPO · CommonsenseQAaccuracy: 73.1 -> 74.9Improved
scale MAGO backbone from 1.5B to 32BMAGO · AIME 2024Pass@1: 0.2741 -> 0.3254Improved
use Pareto-adaptive weights instead of vanilla GRPO (ablation)hybrid reasoning RL · training dynamicsthink-mode shareImproved
use large fixed alpha (0.01) or small alpha (0.0001) (ablation)static-weight DeGRPO · mode selectionmode balance: alpha=0.0001: >90% short; alpha=0.01: >80% thinkHurt

Experimental setup

Builds on: chain-of-thought

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