# Skill execution on SkillsBench Source: [Skill execution on SkillsBench]() Collection index: [llms.txt]() Data updated: 2026-06-02T10:34:38.893Z Evaluation of model performance on general skill execution tasks within the SkillsBench framework, reporting average success rates (avg@5) across various domains like Software Engineering, Office, and Science. 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 | Skill execution | | Dataset (including any stated split) | SkillsBench | | Metric | Overall Success Rate (avg@5) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 26 of 26 | | 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 Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [SW Eng Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Office Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Science Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Media Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Finance Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Cyber Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Robotics Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Other Success Rate (avg@5)](): Higher is better; unit: PERCENT. - [Overall Success Rate Change](): Higher is better; unit: SCALAR. ## Recorded results | Method | Overall Success Rate (avg@5) | Result date | Report paper | | --- | --- | --- | --- | | Evolver | 56.9 | Not recorded | 6a02b47edbc836b87d824c68 | | Revision v3 | 53 | Not recorded | 6a1e6403dbd7150506f855be | | Evolver | 48.2 | Not recorded | 6a02b47edbc836b87d824c68 | | Revision v3 | 44 | Not recorded | 6a1e6403dbd7150506f855be | | Human-curated skill | 43.6 | Not recorded | 6a02b47edbc836b87d824c68 | | Revision v3 | 41 | Not recorded | 6a1e6403dbd7150506f855be | | Revision v3 | 37 | Not recorded | 6a1e6403dbd7150506f855be | | Skill v0 | 35 | Not recorded | 6a1e6403dbd7150506f855be | | Skill-Creator | 34 | Not recorded | 6a1e6403dbd7150506f855be | | SkillCreator-SkillsBench | 33.9 | Not recorded | 6a02b47edbc836b87d824c68 | | Self-Gen | 32 | Not recorded | 6a02b47edbc836b87d824c68 | | No skill | 31 | Not recorded | 6a1e6403dbd7150506f855be | | No skill | 29.9 | Not recorded | 6a02b47edbc836b87d824c68 | | Skill-Creator | 28 | Not recorded | 6a1e6403dbd7150506f855be | | Revision v3 | 27 | Not recorded | 6a1e6403dbd7150506f855be | | Skill v0 | 23 | Not recorded | 6a1e6403dbd7150506f855be | | Skill-Creator | 21 | Not recorded | 6a1e6403dbd7150506f855be | | Skill v0 | 19 | Not recorded | 6a1e6403dbd7150506f855be | | Skill v0 | 17 | Not recorded | 6a1e6403dbd7150506f855be | | Skill-Creator | 17 | Not recorded | 6a1e6403dbd7150506f855be | | No skill | 16 | Not recorded | 6a1e6403dbd7150506f855be | | No skill | 14 | Not recorded | 6a1e6403dbd7150506f855be | | No skill | 11 | Not recorded | 6a1e6403dbd7150506f855be | | Skill v0 | 9 | Not recorded | 6a1e6403dbd7150506f855be | | Skill-Creator | 8 | Not recorded | 6a1e6403dbd7150506f855be | | No skill | 6 | Not recorded | 6a1e6403dbd7150506f855be | ### Evolver Overall Success Rate (avg@5): 56.9. Result ID: 6a034b9647f34ce924e88bd4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Refinement iterations (R)":"2"}. This record does not assert independent reproduction. ### Revision v3 Overall Success Rate (avg@5): 53. Result ID: 6a1eb1a3fbc373592c198c6b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.5","Harness":"Codex"}. This record does not assert independent reproduction. ### Evolver Overall Success Rate (avg@5): 48.2. Result ID: 6a034b9647f34ce924e88bd3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Refinement iterations (R)":"1"}. This record does not assert independent reproduction. ### Revision v3 Overall Success Rate (avg@5): 44. Result ID: 6a1eb1a3fbc373592c198c6f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Opus-4.7","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Human-curated skill Overall Success Rate (avg@5): 43.6. Result ID: 6a034b9647f34ce924e88bd0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Revision v3 Overall Success Rate (avg@5): 41. Result ID: 6a1eb1a3fbc373592c198c7b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepSeek-V4-Pro","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Revision v3 Overall Success Rate (avg@5): 37. Result ID: 6a1eb1a3fbc373592c198c73. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Kimi-2.6","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill v0 Overall Success Rate (avg@5): 35. Result ID: 6a1eb1a3fbc373592c198c6a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"GPT-5.5","Harness":"Codex"}. This record does not assert independent reproduction. ### Skill-Creator Overall Success Rate (avg@5): 34. Result ID: 6a1eb1a3fbc373592c198c69. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.5","Harness":"Codex"}. This record does not assert independent reproduction. ### SkillCreator-SkillsBench Overall Success Rate (avg@5): 33.9. Result ID: 6a034b9647f34ce924e88bd2. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Self-Gen Overall Success Rate (avg@5): 32. Result ID: 6a034b9647f34ce924e88bd1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### No skill Overall Success Rate (avg@5): 31. Result ID: 6a1eb1a3fbc373592c198c68. [Reporting paper]() Recorded evaluation configuration: {"Model":"GPT-5.5","Harness":"Codex"}. This record does not assert independent reproduction. ### No skill Overall Success Rate (avg@5): 29.9. Result ID: 6a034b9647f34ce924e88bcf. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Skill-Creator Overall Success Rate (avg@5): 28. Result ID: 6a1eb1a3fbc373592c198c6d. [Reporting paper]() Recorded evaluation configuration: {"Model":"Opus-4.7","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Revision v3 Overall Success Rate (avg@5): 27. Result ID: 6a1eb1a3fbc373592c198c77. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen-3.6plus","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill v0 Overall Success Rate (avg@5): 23. Result ID: 6a1eb1a3fbc373592c198c7a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepSeek-V4-Pro","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill-Creator Overall Success Rate (avg@5): 21. Result ID: 6a1eb1a3fbc373592c198c79. [Reporting paper]() Recorded evaluation configuration: {"Model":"DeepSeek-V4-Pro","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill v0 Overall Success Rate (avg@5): 19. Result ID: 6a1eb1a3fbc373592c198c72. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Kimi-2.6","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill v0 Overall Success Rate (avg@5): 17. Result ID: 6a1eb1a3fbc373592c198c6e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Opus-4.7","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill-Creator Overall Success Rate (avg@5): 17. Result ID: 6a1eb1a3fbc373592c198c71. [Reporting paper]() Recorded evaluation configuration: {"Model":"Kimi-2.6","Harness":"Claude Code"}. This record does not assert independent reproduction. ### No skill Overall Success Rate (avg@5): 16. Result ID: 6a1eb1a3fbc373592c198c6c. [Reporting paper]() Recorded evaluation configuration: {"Model":"Opus-4.7","Harness":"Claude Code"}. This record does not assert independent reproduction. ### No skill Overall Success Rate (avg@5): 14. Result ID: 6a1eb1a3fbc373592c198c78. [Reporting paper]() Recorded evaluation configuration: {"Model":"DeepSeek-V4-Pro","Harness":"Claude Code"}. This record does not assert independent reproduction. ### No skill Overall Success Rate (avg@5): 11. Result ID: 6a1eb1a3fbc373592c198c70. [Reporting paper]() Recorded evaluation configuration: {"Model":"Kimi-2.6","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill v0 Overall Success Rate (avg@5): 9. Result ID: 6a1eb1a3fbc373592c198c76. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen-3.6plus","Harness":"Claude Code"}. This record does not assert independent reproduction. ### Skill-Creator Overall Success Rate (avg@5): 8. Result ID: 6a1eb1a3fbc373592c198c75. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen-3.6plus","Harness":"Claude Code"}. This record does not assert independent reproduction. ### No skill Overall Success Rate (avg@5): 6. Result ID: 6a1eb1a3fbc373592c198c74. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen-3.6plus","Harness":"Claude Code"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.