AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

SyncLoop: A Multimodal Dual-Loop Framework for Self-Improving Mathematical Reasoning

xiuwei chen, Wentao Hu, Hanhui Li, Yongxin Wang, Jun Zhou, Zisheng Chen, Meng Cao, Yihan Zeng, Kui Zhang, Yu-Jie Yuan, Jianhua Han, Hang Xu and 1 more

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 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
run three SyncLoop iterationsQwen2-VL-7B · Geo-Subaccuracy: 40.4 -> 50.9Improved
run three SyncLoop iterationsQwen2-VL-2B · Geo-Subaccuracy: 28.0 -> 38.2Improved
run three SyncLoop iterationsQwen2-VL-7B · MathVista (ALL, testmini)accuracy: 60.0 -> 63.2Improved
add SFT before RL in second iteration (vs RL-only from pi^1) (ablation)SyncLoop · Geo-Sub-Auxaccuracy: 50.6 -> 40.7Hurt
add SFT before RL in third iteration (vs RL-only from pi^2) (ablation)SyncLoop · Geo-Sub-Auxaccuracy: 54.9 -> 42.9Hurt
warm-start from previous iteration model instead of restarting from initial model (RL only) (ablation)SyncLoop · Geo-Sub-Auxaccuracy: 44.0 -> 50.6Improved
add SFT+RL vs RL only at first iteration (ablation)SyncLoop · Geo-Sub-Auxaccuracy: 44.7 -> 48.7Improved
add SFT to RL when restarting from initial model (ablation)SyncLoop · Geo-Sub-Auxaccuracy: 44.0 -> 48.4Improved
complexify images in addition to text (ablation)SyncLoop iter 3 · Geo-Sub-Auxaccuracy: 51.27 -> 54.9Improved
complexify images in addition to text (ablation)SyncLoop iter 2 · Geo-Sub-Auxaccuracy: 45.82 -> 50.9Improved
filter by error rate >= 0.3 vs using full dataset (ablation)SyncLoop iter 2 · Geo-Sub-Auxaccuracy: 48.4/46.6 -> 49.1/50.9Mixed
raise error-rate threshold to 1.0 (only hardest samples) (ablation)SyncLoop iter 2 · Geo-Sub-Auxaccuracy: 49.1 -> 46.2Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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