# Controllable Writing on WritingBench (WB) Source: [Controllable Writing on WritingBench (WB)]() Collection index: [llms.txt]() Data updated: 2026-04-14T12:11:55.484Z Evaluation of model performance on the WritingBench (WB) dataset focusing specifically on controllable writing capabilities, measured across metrics WB-A through WB-F. 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 | Controllable writing | | Dataset (including any stated split) | WritingBench (WB) | | Metric | WB-A Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [WB-A Score](): Higher is better; unit: PERCENT. - [WB-B Score](): Higher is better; unit: PERCENT. - [WB-C Score](): Higher is better; unit: PERCENT. - [WB-D Score](): Higher is better; unit: PERCENT. - [WB-E Score](): Higher is better; unit: PERCENT. - [WB-F Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | WB-A Score | Result date | Report paper | | --- | --- | --- | --- | | Qwen3-32B | 79.8 | Not recorded | 69ddca787d0172140ec79f31 | | Claude 3.7 | 78.2 | Not recorded | 69ddca787d0172140ec79f31 | | DeepWriter-8B + RL-TCER | 76.5 | Not recorded | 69ddca787d0172140ec79f31 | | DeepWriter-8B + RL-EndoR | 76 | Not recorded | 69ddca787d0172140ec79f31 | | GPT-4o | 74.4 | Not recorded | 69ddca787d0172140ec79f31 | | Qwen2.5-7B-Instruct + SFT + RL-TCER | 72.7 | Not recorded | 69ddca787d0172140ec79f31 | | Qwen2.5-7B-Instruct + SFT + RL-EndoR | 72.6 | Not recorded | 69ddca787d0172140ec79f31 | | DeepWriter-8B | 72.2 | Not recorded | 69ddca787d0172140ec79f31 | | Qwen2.5-7B-Instruct + SFT | 69.7 | Not recorded | 69ddca787d0172140ec79f31 | | Qwen3-8B | 68.7 | Not recorded | 69ddca787d0172140ec79f31 | | LongWriter-8B + RL-TCER | 65.8 | Not recorded | 69ddca787d0172140ec79f31 | | LongWriter-8B + RL-EndoR | 64.4 | Not recorded | 69ddca787d0172140ec79f31 | | Claude 3.5 | 59.5 | Not recorded | 69ddca787d0172140ec79f31 | | Qwen2.5-7B-Instruct | 58.9 | Not recorded | 69ddca787d0172140ec79f31 | | LongWriter-8B | 57.9 | Not recorded | 69ddca787d0172140ec79f31 | | Qwen2.5-32B-Instruct | 52.5 | Not recorded | 69ddca787d0172140ec79f31 | | Llama3.1-8B-Instruct | 47.5 | Not recorded | 69ddca787d0172140ec79f31 | ### Qwen3-32B WB-A Score: 79.8. Result ID: 69de2eec47f34ce924c6eb23. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction. ### Claude 3.7 WB-A Score: 78.2. Result ID: 69de2eec47f34ce924c6eb1f. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Closed-source Models"}. This record does not assert independent reproduction. ### DeepWriter-8B + RL-TCER WB-A Score: 76.5. Result ID: 69de2eec47f34ce924c6eb03. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Source Finetuning Dataset":"DeepWriting Dataset"}. This record does not assert independent reproduction. ### DeepWriter-8B + RL-EndoR WB-A Score: 76. Result ID: 69de2eec47f34ce924c6eb02. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Source Finetuning Dataset":"DeepWriting Dataset"}. This record does not assert independent reproduction. ### GPT-4o WB-A Score: 74.4. Result ID: 69de2eec47f34ce924c6eb1d. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Closed-source Models"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct + SFT + RL-TCER WB-A Score: 72.7. Result ID: 69de2eec47f34ce924c6eb27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"SFT + RL-TCER"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct + SFT + RL-EndoR WB-A Score: 72.6. Result ID: 69de2eec47f34ce924c6eb26. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"SFT + RL-EndoR"}. This record does not assert independent reproduction. ### DeepWriter-8B WB-A Score: 72.2. Result ID: 69de2eec47f34ce924c6eb01. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Source Finetuning Dataset":"DeepWriting Dataset"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct + SFT WB-A Score: 69.7. Result ID: 69de2eec47f34ce924c6eb25. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"SFT"}. This record does not assert independent reproduction. ### Qwen3-8B WB-A Score: 68.7. Result ID: 69de2eec47f34ce924c6eb22. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction. ### LongWriter-8B + RL-TCER WB-A Score: 65.8. Result ID: 69de2eec47f34ce924c6eb06. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Source Finetuning Dataset":"LongWriter Dataset"}. This record does not assert independent reproduction. ### LongWriter-8B + RL-EndoR WB-A Score: 64.4. Result ID: 69de2eec47f34ce924c6eb05. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Source Finetuning Dataset":"LongWriter Dataset"}. This record does not assert independent reproduction. ### Claude 3.5 WB-A Score: 59.5. Result ID: 69de2eec47f34ce924c6eb1e. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Closed-source Models"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct WB-A Score: 58.9. Result ID: 69de2eec47f34ce924c6eb24. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"Base"}. This record does not assert independent reproduction. ### LongWriter-8B WB-A Score: 57.9. Result ID: 69de2eec47f34ce924c6eb04. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Source Finetuning Dataset":"LongWriter Dataset"}. This record does not assert independent reproduction. ### Qwen2.5-32B-Instruct WB-A Score: 52.5. Result ID: 69de2eec47f34ce924c6eb21. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction. ### Llama3.1-8B-Instruct WB-A Score: 47.5. Result ID: 69de2eec47f34ce924c6eb20. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.