TIR-Bench: Results as Published

TIR-Bench tests agentic visual reasoning in which the model must use tools to change an image while it thinks, rather than answer from one look. It has 13 tasks that require operations such as zooming, rotating, raising contrast, drawing auxiliary lines, cropping, flipping, or reassembling a jigsaw. The paper scores accuracy, and it also checks whether a model calls the right tool with the right parameters on a rotation task. A code interpreter is one tool setting in the experiments, not a separate definition of the benchmark.

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.