arXiv:2601.21288v2 Announce Type: replace Abstract: Autonomous driving is an important and safety-critical task, and recent advances in LLMs/VLMs have opened new possibilities for reasoning and planning in this domain. However, large models demand substantial GPU memory and exhibit high inference latency, while conventional supervised fine-tuning (SFT) often struggles to bridge the capability gaps of small models. To address these limitations, we propose Drive-KD, a framework that decomposes autonomous driving into a "perception-reasoning-planning" triad and transfers these capabilities via kn
The proliferation of powerful LLMs/VLMs is driving innovation in complex domains like autonomous driving, necessitating efficient deployment strategies.
This research addresses critical limitations of large models in autonomous driving by reducing computational demands and increasing accessibility, which is crucial for safety-critical applications.
The proposed Drive-KD framework potentially allows smaller, more efficient models to achieve capabilities closer to large VLMs in autonomous driving through multi-teacher distillation.
- · Autonomous driving companies
- · Edge AI hardware developers
- · Developers of efficient AI models
- · Developers reliant solely on large, computationally intensive models
Improved performance and reduced cost for autonomous driving systems.
Accelerated adoption of advanced AI in vehicles due to lower barriers to entry for smaller models.
Enhanced safety and reliability of autonomous vehicles leading to broader public acceptance and regulatory shifts.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI