Natural Plan
A planning evaluation benchmark for language models proposed by Google. It includes real-world scenario tasks such as trip scheduling, meeting arrangement, and calendar planning, requiring models to generate multi-step plans satisfying multiple constraints, with correctness verified programmatically.
Timeline (last 90 days)
On the Natural Plan benchmark without task-specific training, the diffusion model's plan repair success rate was 53.7%, while the autoregressive model's was 27.0%.
Natural Plan is a benchmark dataset proposed by Google for evaluating the planning capabilities of language models, including tasks such as itinerary scheduling, meeting scheduling, and calendar planning.
All Facts (2)
On the Natural Plan benchmark without task-specific training, the diffusion model's plan repair success rate was 53.7%, while the autoregressive model's was 27.0%.
50%UnverifiedNatural Plan is a benchmark dataset proposed by Google for evaluating the planning capabilities of language models, including tasks such as itinerary scheduling, meeting scheduling, and calendar planning.
50%