# Multi-Turn Tool Calling on BFCL v4 (val) Source: [Multi-Turn Tool Calling on BFCL v4 (val)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:54:25.382Z Evaluation of multi-turn tool calling capabilities using the validation split of the BFCL v4 dataset, reporting metrics related to overall accuracy, base performance, function/parameter misses, and long context handling. 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 | Multi-Turn Tool Calling | | Dataset (including any stated split) | BFCL v4 (val) | | Metric | Overall Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 15 of 15 | | 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 - [Overall Accuracy](): Higher is better; unit: PERCENT. - [Base Performance](): Higher is better; unit: PERCENT. - [Function Miss Rate](): Lower is better; unit: PERCENT. - [Parameter Miss Rate](): Lower is better; unit: PERCENT. - [Long Context Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Overall Accuracy | Result date | Report paper | | --- | --- | --- | --- | | RCTP-FT + RC-GRPO | 85 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | RCTP-FT + GRPO | 73.75 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | Opus-4.5 | 61.25 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | GLM-4.7 | 58.75 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | Sonnet-4.5 | 57.5 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | Gemini-3-Pro | 53.75 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | GPT-5.2 | 50 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | SFT + GRPO | 48.75 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | RCTP-FT + RC-GRPO | 48.75 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | SFT + RC-GRPO | 46.25 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | RCTP-FT + GRPO | 46.25 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | SFT + GRPO | 35 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | SFT + RC-GRPO | 35 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | Base Model | 11.25 | Not recorded | 6982d2e1a8b01e76cc8e9861 | | Base Model | 0 | Not recorded | 6982d2e1a8b01e76cc8e9861 | ### RCTP-FT + RC-GRPO Overall Accuracy: 85. Result ID: 699608a55698665ec44048ad. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### RCTP-FT + GRPO Overall Accuracy: 73.75. Result ID: 699608a55698665ec44048ac. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### Opus-4.5 Overall Accuracy: 61.25. Result ID: 699608a55698665ec44048b3. [Reporting paper]() Recorded evaluation configuration: {"Version":"claude-opus-4-5-20251101"}. This record does not assert independent reproduction. ### GLM-4.7 Overall Accuracy: 58.75. Result ID: 699608a55698665ec44048b4. [Reporting paper]() Recorded evaluation configuration: {"Version":"glm-4.7"}. This record does not assert independent reproduction. ### Sonnet-4.5 Overall Accuracy: 57.5. Result ID: 699608a55698665ec44048b5. [Reporting paper]() Recorded evaluation configuration: {"Version":"claude-sonnet-4-5-20250929"}. This record does not assert independent reproduction. ### Gemini-3-Pro Overall Accuracy: 53.75. Result ID: 699608a55698665ec44048b6. [Reporting paper]() Recorded evaluation configuration: {"Version":"gemini-3-pro-preview"}. This record does not assert independent reproduction. ### GPT-5.2 Overall Accuracy: 50. Result ID: 699608a55698665ec44048b7. [Reporting paper]() Recorded evaluation configuration: {"Version":"gpt-5.2-2025-12-11"}. This record does not assert independent reproduction. ### SFT + GRPO Overall Accuracy: 48.75. Result ID: 699608a55698665ec44048aa. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### RCTP-FT + RC-GRPO Overall Accuracy: 48.75. Result ID: 699608a55698665ec44048b2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B"}. This record does not assert independent reproduction. ### SFT + RC-GRPO Overall Accuracy: 46.25. Result ID: 699608a55698665ec44048ab. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### RCTP-FT + GRPO Overall Accuracy: 46.25. Result ID: 699608a55698665ec44048b1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B"}. This record does not assert independent reproduction. ### SFT + GRPO Overall Accuracy: 35. Result ID: 699608a55698665ec44048af. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B"}. This record does not assert independent reproduction. ### SFT + RC-GRPO Overall Accuracy: 35. Result ID: 699608a55698665ec44048b0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B"}. This record does not assert independent reproduction. ### Base Model Overall Accuracy: 11.25. Result ID: 699608a55698665ec44048a9. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### Base Model Overall Accuracy: 0. Result ID: 699608a55698665ec44048ae. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.