Tag
This paper presents a method for learning implicit causal world models from multi-agent demonstrations, enabling agents to infer causal structures from observed behavior.
The article discusses how AI agents are being widely demonstrated for tasks like browsing, coding, and automating workflows, but their commercialization remains immature, with unclear answers about who pays and how.
This paper reveals a counterintuitive phenomenon where correct demonstrations in in-context learning can degrade model accuracy, introducing task preserving perturbations to study the gap between exemplar correctness and utility.
Introduces Self-Distillation Fine-Tuning (SDFT), a method that enables on-policy learning from demonstrations to achieve continual learning without catastrophic forgetting, outperforming supervised fine-tuning.