CL-bench is a benchmark dataset for evaluating the context learning capabilities of a large language model, released in 2026 by Tencent's Hunyuan team in collaboration with Fudan University. The related research papers are as follows: CL-bench: A Benchmark for Context LearningThe aim is to test whether a model can learn new rules, concepts, or domain knowledge from a given context without relying on pre-trained knowledge and apply them to subsequent tasks.

This dataset contains 500 complex context scenarios, covering 1,899 specific tasks, and provides 31,607 fine-grained evaluation rubrics. Each task is organized in a multi-turn dialogue format, covering various context learning scenarios such as rule reasoning, domain knowledge learning, and complex instruction understanding, to evaluate the model's ability to understand, summarize, and transfer new information in the context.