Before Thinking, Learn to Decide: Proactive Routing for Efficient Visual Reasoning
Large multimodal models have achieved strong reasoning on complex visual tasks, but their inference efficiency is often restricted by long chains of thought. A promising solution is to pair a small draft model with a large target model, enabling cooperative inference employing a routing signal that adaptively routes queries to either the draft or target model based on their difficulties for optimal efficiency and accuracy. Yet, the remaining bottleneck is to establish a reliable query difficulty signal under multimodal settings. Existing approaches designed for language models either rely on p
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- LinkedLinked via unknownNew benchmark exposes reasoning gaps in top models →
- LinkedLinked via unknownVioletVision-3B →
- FuzzySimilar title/name (fuzzy) · 59%rasbt/reasoning-from-scratch →
“Fuzzy title match (0.73): “Before Thinking, Learn to Decide: Proactive Routing for Effi” ≈ “rasbt/reasoning-from-scratch””
- FuzzySimilar title/name (fuzzy) · 59%scikit-learn/scikit-learn →
“Fuzzy title match (0.73): “Before Thinking, Learn to Decide: Proactive Routing for Effi” ≈ “scikit-learn/scikit-learn””
- FuzzySimilar title/name (fuzzy) · 59%shareAI-lab/learn-claude-code →
“Fuzzy title match (0.73): “Before Thinking, Learn to Decide: Proactive Routing for Effi” ≈ “shareAI-lab/learn-claude-code””