# Long-form Writing on WritingBench Source: [Long-form Writing on WritingBench]() Collection index: [llms.txt]() Data updated: 2026-04-21T08:47:02.086Z Evaluation of long-form writing performance on the WritingBench dataset, measured using the general 'Score' metric. 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 | Long-form Writing | | Dataset (including any stated split) | WritingBench | | Metric | Score | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 18 of 18 | | 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 - [Score](): Higher is better; unit: SCALAR. ## Recorded results | Method | Score | Result date | Report paper | | --- | --- | --- | --- | | Qwen2.5-14B-Writing-RL | 88.27 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-7B-Writing-RL | 87.23 | Not recorded | 68fa792c1e209131cea7c671 | | Llama3.1-8B-Writing-RL | 87.11 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-7B-Writing-RL | 86.29 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-14B-WritingBench-SFT | 84.95 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-7B-Reference-SFT | 84.23 | Not recorded | 68fa792c1e209131cea7c671 | | Llama3.1-8B-Reference-SFT | 83.98 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-7B-WritingBench-SFT (24k) | 83.76 | Not recorded | 68fa792c1e209131cea7c671 | | Llama3.1-8B-WritingBench-SFT | 83.75 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-7B-WritingBench-SFT (12k) | 83.71 | Not recorded | 68fa792c1e209131cea7c671 | | GPT-4o | 83.42 | Not recorded | 68fa792c1e209131cea7c671 | | Longwriter-Zero-32B | 82.92 | Not recorded | 68fa792c1e209131cea7c671 | | Longwriter-9B | 79.1 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen-Plus | 77.62 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-7B-Instruct | 73.16 | Not recorded | 68fa792c1e209131cea7c671 | | Qwen2.5-14B-Instruct | 72.67 | Not recorded | 68fa792c1e209131cea7c671 | | Llama3.1-8B-Instruct | 66.56 | Not recorded | 68fa792c1e209131cea7c671 | | Suri-7B | 49.7 | Not recorded | 68fa792c1e209131cea7c671 | ### Qwen2.5-14B-Writing-RL Score: 88.27. Result ID: 69e7397d47f34ce924cd884f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"PPO"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Writing-RL Score: 87.23. Result ID: 69e7397d47f34ce924cd884b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"PPO"}. This record does not assert independent reproduction. ### Llama3.1-8B-Writing-RL Score: 87.11. Result ID: 69e7397d47f34ce924cd8853. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"PPO"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Writing-RL Score: 86.29. Result ID: 69e7397d47f34ce924cd884c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"GRPO"}. This record does not assert independent reproduction. ### Qwen2.5-14B-WritingBench-SFT Score: 84.95. Result ID: 69e7397d47f34ce924cd884e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Reference-SFT Score: 84.23. Result ID: 69e7397d47f34ce924cd884a. [Reporting paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. ### Llama3.1-8B-Reference-SFT Score: 83.98. Result ID: 69e7397d47f34ce924cd8852. [Reporting paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. ### Qwen2.5-7B-WritingBench-SFT (24k) Score: 83.76. Result ID: 69e7397d47f34ce924cd8849. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. ### Llama3.1-8B-WritingBench-SFT Score: 83.75. Result ID: 69e7397d47f34ce924cd8851. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. ### Qwen2.5-7B-WritingBench-SFT (12k) Score: 83.71. Result ID: 69e7397d47f34ce924cd8848. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. ### GPT-4o Score: 83.42. Result ID: 69e7397d47f34ce924cd8843. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"–","RL":"–"}. This record does not assert independent reproduction. ### Longwriter-Zero-32B Score: 82.92. Result ID: 69e7397d47f34ce924cd8846. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"GRPO"}. This record does not assert independent reproduction. ### Longwriter-9B Score: 79.1. Result ID: 69e7397d47f34ce924cd8845. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"DPO"}. This record does not assert independent reproduction. ### Qwen-Plus Score: 77.62. Result ID: 69e7397d47f34ce924cd8842. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"–","RL":"–"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct Score: 73.16. Result ID: 69e7397d47f34ce924cd8847. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✗","RL":"✗"}. This record does not assert independent reproduction. ### Qwen2.5-14B-Instruct Score: 72.67. Result ID: 69e7397d47f34ce924cd884d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✗","RL":"✗"}. This record does not assert independent reproduction. ### Llama3.1-8B-Instruct Score: 66.56. Result ID: 69e7397d47f34ce924cd8850. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"SFT":"✗","RL":"✗"}. This record does not assert independent reproduction. ### Suri-7B Score: 49.7. Result ID: 69e7397d47f34ce924cd8844. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"SFT":"✓","RL":"✗"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.