# Per-step Tool Selection Prediction on τ-bench airline multi-step Source: [Per-step Tool Selection Prediction on τ-bench airline multi-step]() Collection index: [llms.txt]() Data updated: 2026-06-16T09:58:01.819Z Evaluation of a model's ability to predict the correct tool selection at each sequential step within the multi-step configuration of the τ-bench airline dataset. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | per-step tool selection prediction | | Dataset (including any stated split) | τ-bench airline multi-step | | Metric | Success Rate | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 5 of 5 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Success Rate](): Higher is better; unit: PERCENT. - [Attention Margin AUROC](): Higher is better; unit: SCALAR. - [Gold Standard Metric (pgold)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Success Rate | Result date | Report paper | | --- | --- | --- | --- | | Qwen2.5-3B | 100 | Not recorded | 6a30d90ad33dfb23e783d1fc | | Llama-3.1-8B | 100 | Not recorded | 6a30d90ad33dfb23e783d1fc | | Qwen2.5-7B | 95 | Not recorded | 6a30d90ad33dfb23e783d1fc | | Qwen2.5-0.5B | 55 | Not recorded | 6a30d90ad33dfb23e783d1fc | | Qwen2.5-1.5B | 49 | Not recorded | 6a30d90ad33dfb23e783d1fc | ### Qwen2.5-3B Success Rate: 100. Result ID: 6a311e14fbc373592c2df8f6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"status":"ceiling"}. This record does not assert independent reproduction. ### Llama-3.1-8B Success Rate: 100. Result ID: 6a311e14fbc373592c2df8f8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"status":"0 fails"}. This record does not assert independent reproduction. ### Qwen2.5-7B Success Rate: 95. Result ID: 6a311e14fbc373592c2df8f7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"status":"noise"}. This record does not assert independent reproduction. ### Qwen2.5-0.5B Success Rate: 55. Result ID: 6a311e14fbc373592c2df8f4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"status":"inform."}. This record does not assert independent reproduction. ### Qwen2.5-1.5B Success Rate: 49. Result ID: 6a311e14fbc373592c2df8f5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"status":"inform."}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.