# agenteval-long-context

Long-context benchmark data (RULER 64K/128K/256K, generated with the GLM-5.3 tokenizer, seed 42, 500 samples/subtask) for the AgentEval long-context suite.