A personal benchmark for evaluating ultra-long-context (million-token) comprehension using long-form web-novel reading comprehension as the task.
Source data & novel on GitHub ↗To avoid spoilers, the novel title and problems are not shown on this page.
Open GitHub repository ↗