AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Training-Free Multimodal Large Language Model Orchestration

Tianyu Xie, Yuexiao Ma, Yuhang Wu, Wang Chen, Jiayi Ji, Tat-Seng Chua, Xiawu Zheng, Rongrong Ji

ICML 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multimodal fusion with missing modalities

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 orchestration over Qwen2-VL on long videosQwen2-VL · Video-MME (long)accuracy: 56.3 -> 69.2Improved
add orchestration over Qwen2-VL overallQwen2-VL · Video-MME (overall)accuracy: 64.9 -> 71.9Improved
add orchestration over LLaVA-Video on long videosLLaVA-Video · Video-MME (long)accuracy: 58.7 -> 72.6Improved
add orchestration over LLaVA-Video overallLLaVA-Video · Video-MME (overall)accuracy: 68.6 -> 76.4Improved
add orchestration over Qwen2-VL on short videosQwen2-VL · Video-MME (short)accuracy: 75.0 -> 76.7Improved
upgrade expert pair from Qwen2.5-VL-32B/72B to Qwen3-VL-32B/235B-A22B (ablation)orchestration with fixed controller · Video-MMEaccuracy: 65.58 -> 72.31Improved
enable evidence-keyed cross-modal memory (ablation)orchestration framework · dialogue sessionsexpert calls: 49% reductionImproved
replace the controller under a fixed token protocol (ablation)orchestration framework · Video-MMEaccuracyNo clear effect
use training-free controller instead of supervised LoRA-trained router (ablation)routing · Video-MME + WorldSensevalid control-token success (%): 99.20 -> 99.10No clear effect
increase average active expert params through routing from 19.7B to 53.8B (ablation)Ours (Qwen2.5-VL) · Video-MMEaccuracy: 62.07 -> 65.58Improved
orchestrate experts instead of dense 72B modelOurs vs Qwen2.5-VL-72B · Video-MMEaccuracy: 65.74 vs 65.58No clear effect
memory reuse in round 2 of three-round dialogueOurs · Video-MME interactiveaccuracy: 66.70 -> 69.20 -> 64.30Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 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