AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Learning to Inference Adaptively for Multimodal Large Language Models

Zhuoyan Xu, Khoi Duc Nguyen, Preeti Mukherjee, Saurabh Bagchi, Somali Chaterji, Yingyu Liang, Yin Li

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 11 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 probabilistic scheduler (sample exactly k switches) instead of deterministic hinge-penalty scheduler (ablation)AdaLLaVA-L 7B · VQAv2accuracy at budget 0.65: 72.2 -> 72.7Improved
use probabilistic scheduler instead of deterministic for budget utilization (ablation)AdaLLaVA-L 7B · VQAv2utilization (%) at budget 0.75: 83.2 -> 100.0Improved
use probabilistic scheduler instead of deterministic for latency constraint success at 0.95 budget (ablation)AdaLLaVA-L 7B · VQAv2success rate (%): 96.1 -> 100.0Improved
apply 24 switches instead of 16 switches (last 16 layers) (ablation)AdaLLaVA-L · VQAv2accuracyHurt
use 4-head instead of 8-head attention switch groups (ablation)AdaLLaVA-H · VQAv2accuracy-latency tradeoffNo clear effect
use head/neuron-level switches (H) instead of layer-level switches (L) (ablation)AdaLLaVA · VQAv2accuracy: about 75% -> about 76%Improved
train the scheduler instead of random uniform sampling of execution plans (ablation)AdaLLaVA-L on Mipha-3B · Mipha-3BaccuracyImproved
reduce latency budget from 100% to 60% for AdaLLaVA-LLLaVA-1.5 Vicuna-7B · TextVQAaccuracy: 57.0 -> 47.7Hurt
reduce latency budget from 100% to 60% for AdaLLaVA-LLLaVA-1.5 Vicuna-7B · MMBenchaccuracy: 63.7 -> 63.8No clear effect
reduce latency budget to 60% on MM-Vet and LLaVA-WILDAdaLLaVA-L-7B · MM-Vet / LLaVA-WILDscore: MM-Vet 30.7 -> 23.9; LLaVA-WILD 64.2 -> 46.3Hurt
integrate AdaLLaVA-H with PruMerge+ token selectionPruMerge+ · VQAv2accuracy and FLOPs: 76.8 -> 76.0 at 100% budget; FLOPs 3.0T at 100% to 1.8T at 60%Mixed
apply AdaLLaVA-L to Mipha-3B at 60% budgetMipha-3B · MMBenchaccuracy: 69.7 -> 64.6Hurt

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