Research Map

Use this page to decide where new material belongs.

Layer 1: Application domain

• Autonomous Driving
• Robotics
• Machine Learning

Layer 2: Stack component

• Perception
• Prediction
• Planning

Layer 3: Paradigm

• End To End Architectures
• Vision Language Action
• Foundation Models

Layer 4: Evaluation and deployment questions

For each new source, tag at least:

• open-loop vs closed-loop,
• simulation vs real-world,
• map dependence,
• sensor assumptions,
• action abstraction,
• data regime,
• deployment evidence.

Source programs

Routing guide for new papers

If the paper is about... → Route to:

• Language + driving action → Vla And Driving + tag vla
• Driving perception/prediction/planning (no language) → Autonomous Driving Seminal Papers
• LLM/VLM architecture (not driving-specific) → Llm Seminal Papers
• Foundational ML / Ilya-adjacent → Ilya Top 30
• Generative models (diffusion, flow matching, VAE) → Foundation Models under diffusion models section
• Robotics VLA → Vla And Driving under general VLA section

VLA sub-taxonomy (from AutoVLA analysis)

Papers in the VLA-driving space can be classified along these axes:

Language Role:        supervision ←→ reasoning ←→ runtime control ←→ action output
Action Space:         controls ←→ waypoints ←→ planner tokens ←→ language tokens
Architecture:         VLM+planner ←→ decoupled ←→ true VLA ←→ MoE
Evaluation:           open-loop ←→ closed-loop sim ←→ real-world
Training:             IL only ←→ IL+SFT ←→ IL+RL ←→ GRPO