AA-LCR: Results as Published

AA-LCR tests whether a model can reason across several long documents at once, rather than retrieve a single passage. The benchmark has 100 human-crafted questions, each tied to a document set averaging about 100,000 tokens (cl100k_base tokenizer), and each answer has to be inferred by combining information from several documents. Undergraduate students wrote the questions, and every question was answered correctly by at least one human tester. The published metric is accuracy, shown as a percentage on the Artificial Analysis leaderboard.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.