AI Paper Atlas

Research map / Large language models / Agents, code and math

Reasoning Boosts Opinion Alignment in LLMs

Frédéric Berdoz, Yann Billeter, Yann Vonlanthen, Roger Wattenhofer

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 0 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 SFT warm start before GRPO (ablation)GRPO, Llama 3.1 8B · WoMmacro-F1: 37.29 -> 52.53Improved
add SFT warm start before GRPO (ablation)GRPO, Llama 3.1 8B · smartvotemacro-F1: 55.14 -> 66.88Improved
add SFT warm start before GRPO (ablation)GRPO, Qwen3 8B · ANESmacro-F1: 31.47 -> 38.44Improved
add SFT warm start before GRPO (ablation)GRPO, Magistral 24B · smartvotemacro-F1: 60.56 -> 70.73Improved
add GRPO after SFT (ablation)SFT, Llama 3.1 8B · ANESmacro-F1: 42.77 -> 40.66Hurt
add GRPO after SFT (ablation)SFT, Llama 3.1 8B · WoMmacro-F1: 48.95 -> 52.53Improved
add GRPO after SFT (ablation)SFT, Magistral 24B · WoMmacro-F1: 51.86 -> 53.21No clear effect
use reasoning-pretrained backbone (Qwen3 8B vs Llama 3.1 8B)SFT+GRPO · WoMmacro-F1: 52.53 -> 49.38Hurt
scale backbone to Magistral 24BSFT+GRPO · smartvotemacro-F1: 66.88 -> 70.73Improved
replace ICL with SFT+GRPOLlama 3.1 8B · WoMmacro-F1: 28.17 -> 52.53Improved
remove Neutral instances from ANES evaluation (ablation)SFT+GRPO, Llama 3.1 8B · ANESaccuracy: 59.21 -> 71.35Improved
use aggressive instead of conservative ANES recoding (ablation)SFT+GRPO, Llama 3.1 8B · ANESmacro-F1: 40.67 -> 45.33Improved

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