AI Paper Atlas

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

JudgeBoard: Benchmarking and Enhancing Small Language Models for Reasoning Evaluation

Zhenyu Bi, Gaurav Srivastava, Yang Li, Swastik Roy, Meng Lu, Morteza Ziyadi, Xuan Wang

AAAI 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 2 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
use multi-agent profiled debate (MAJ) instead of single judgeQwen3-14B · MATH Algebraoverall judging accuracy: 0.6684 -> 0.7789Improved
use MAJ instead of single judgeQwen3-14B · MATH Number Theoryoverall judging accuracy: 0.6923 -> 0.9385Improved
use MAJ instead of single judgeQwen3-14B · MATH Counting and Probabilityoverall judging accuracy: 0.7111 -> 0.9056Improved
use MAJ instead of single judgeQwen3-8B · MATH Number Theoryoverall judging accuracy: 0.6308 -> 0.9Improved
use MAJ instead of single judgeQwen3-4B · MATH Algebraoverall judging accuracy: 0.6421 -> 0.7632Improved
use MAJ instead of single judgeGemma3-12b · MATH Algebraoverall judging accuracy: 0.6737 -> 0.7Improved
use MAJ instead of single judgeGemma3-12b · MATH Number Theoryoverall judging accuracy: 0.7923 -> 0.7231Hurt
use MAJ instead of single judgePhi4 · MATH Algebraoverall judging accuracy: 0.6737 -> 0.6421Hurt
use MAJ instead of single judgePhi4 · MATH Number Theoryoverall judging accuracy: 0.7077 -> 0.6692Hurt
enable reasoning mode (ablation)Qwen3-4B · MATH Algebraoverall judging accuracy: 0.6421 -> 0.6474Improved
enable reasoning mode (ablation)Qwen3-4B · MATH Number Theoryoverall judging accuracy: 0.7384 -> 0.7000Hurt
enable reasoning mode (ablation)Qwen3-14B · MATH Number Theoryoverall judging accuracy: 0.6923 -> 0.7077Improved

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