language:
- en
license: cc-by-4.0
configs:
- config_name: prompts
data_files:
- split: latest
path: prompts.jsonl
Dataset Card for Amazon Nova Canvas Evaluation
This document contains some additional context on the settings and methodology for how we evaluated the Nova Canvas model. Unless stated otherwise, we use greedy decoding with the maximum generation length set at 1600 tokens.
Responses from the Nova models are released under a CC-BY-NC license. The public benchmarks used to generate responses can be accessed via the hyperlinks provided below, subject to the applicable license terms for each benchmark
Nova Canvas Human Evaluation Prompts
The A/B testing prompt set is composed of approximately 1,000 prompts designed to capture customer usage of text-to-image models. This set include prompts from datasets such as MSCOCO, Drawbench, OpenParti, DALL.E 3 Eval, and DOCCI and covers a broad set of categories such as humans, landscapes, natural scenarios, indoor environments, creative themes, artistic themes, and so forth. A few prompts were randomly selected and repeated in order to get additional data points on the quality of the model.