3 accepted papers
We introduce Context Tuning, a simple and effective method to significantly enhance few-shot adaptation of language models (LLMs) without fine-tuning model parameters. While prompt-based adaptation techniques have demonstrated the effectiveness of lightweight adaptation methods for LLMs, they typica…
Reasoning models leveraging long chains of thought employ various cognitive skills, such as verification of their answers, backtracking, retrying by an alternate method, and more. Previous work has shown that when a base language model exhibits these skills, training that model further with reinforc…
We consider the task of traffic scene generation. A common approach in the self-driving industry is to use manual creation to generate scenes with specific characteristics and automatic generation to generate canonical scenes at scale. However, manual creation is not scalable, and automatic generati…