# Terminal-based problem solving on Terminal-Bench 2 (out-of-distribution) Source: [Terminal-based problem solving on Terminal-Bench 2 (out-of-distribution)]() Collection index: [llms.txt]() Data updated: 2026-05-26T11:01:55.309Z Evaluation of model performance on solving tasks within the out-of-distribution set of Terminal-Bench 2, focusing on task success rate and average reasoning steps required. 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 | Terminal-based problem solving | | Dataset (including any stated split) | Terminal-Bench 2 (out-of-distribution) | | Metric | Task Success Rate | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 - [Task Success Rate](): Higher is better; unit: PERCENT. - [Average Reasoning Steps](): Lower is better; unit: SCALAR. ## Recorded results | Method | Task Success Rate | Result date | Report paper | | --- | --- | --- | --- | | CODESKILL | 34.12 | Not recorded | 6a152968dbd7150506f7e373 | | Subtask Memory | 30.59 | Not recorded | 6a152968dbd7150506f7e373 | | Prompt Skill Mgmt. | 28.24 | Not recorded | 6a152968dbd7150506f7e373 | | No-skill baseline | 25.88 | Not recorded | 6a152968dbd7150506f7e373 | | CODESKILL | 25.88 | Not recorded | 6a152968dbd7150506f7e373 | | Prompt Skill Mgmt. | 24.71 | Not recorded | 6a152968dbd7150506f7e373 | | SFT-CODESKILL | 24.71 | Not recorded | 6a152968dbd7150506f7e373 | | Prompt Skill Mgmt. | 23.53 | Not recorded | 6a152968dbd7150506f7e373 | | SFT-CODESKILL | 23.53 | Not recorded | 6a152968dbd7150506f7e373 | | Subtask Memory | 22.35 | Not recorded | 6a152968dbd7150506f7e373 | | Prompt Skill Mgmt. | 21.18 | Not recorded | 6a152968dbd7150506f7e373 | | No-skill baseline | 20 | Not recorded | 6a152968dbd7150506f7e373 | ### CODESKILL Task Success Rate: 34.12. Result ID: 6a157da1fbc373592c0dc420. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Skill Backbone":"Qwen3.5-4B","Frozen coding policy":"Qwen3.5-35B-A3B"}. This record does not assert independent reproduction. ### Subtask Memory Task Success Rate: 30.59. Result ID: 6a157da1fbc373592c0dc41c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Skill Backbone":"GPT-5.4-mini","Frozen coding policy":"Qwen3.5-35B-A3B"}. This record does not assert independent reproduction. ### Prompt Skill Mgmt. Task Success Rate: 28.24. Result ID: 6a157da1fbc373592c0dc41e. [Reporting paper]() Recorded evaluation configuration: {"Skill Backbone":"GPT-5.4-mini","Frozen coding policy":"Qwen3.5-35B-A3B"}. This record does not assert independent reproduction. ### No-skill baseline Task Success Rate: 25.88. Result ID: 6a157da1fbc373592c0dc41b. [Reporting paper]() Recorded evaluation configuration: {"Skill Backbone":"-","Frozen coding policy":"Qwen3.5-35B-A3B"}. This record does not assert independent reproduction. ### CODESKILL Task Success Rate: 25.88. Result ID: 6a157da1fbc373592c0dc426. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Skill Backbone":"Qwen3.5-4B","Frozen coding policy":"GPT-5.4-mini"}. This record does not assert independent reproduction. ### Prompt Skill Mgmt. Task Success Rate: 24.71. Result ID: 6a157da1fbc373592c0dc41d. [Reporting paper]() Recorded evaluation configuration: {"Skill Backbone":"Qwen3.5-4B","Frozen coding policy":"Qwen3.5-35B-A3B"}. This record does not assert independent reproduction. ### SFT-CODESKILL Task Success Rate: 24.71. Result ID: 6a157da1fbc373592c0dc41f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Skill Backbone":"Qwen3.5-4B","Frozen coding policy":"Qwen3.5-35B-A3B"}. This record does not assert independent reproduction. ### Prompt Skill Mgmt. Task Success Rate: 23.53. Result ID: 6a157da1fbc373592c0dc423. [Reporting paper]() Recorded evaluation configuration: {"Skill Backbone":"Qwen3.5-4B","Frozen coding policy":"GPT-5.4-mini"}. This record does not assert independent reproduction. ### SFT-CODESKILL Task Success Rate: 23.53. Result ID: 6a157da1fbc373592c0dc425. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Skill Backbone":"Qwen3.5-4B","Frozen coding policy":"GPT-5.4-mini"}. This record does not assert independent reproduction. ### Subtask Memory Task Success Rate: 22.35. Result ID: 6a157da1fbc373592c0dc422. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Skill Backbone":"GPT-5.4-mini","Frozen coding policy":"GPT-5.4-mini"}. This record does not assert independent reproduction. ### Prompt Skill Mgmt. Task Success Rate: 21.18. Result ID: 6a157da1fbc373592c0dc424. [Reporting paper]() Recorded evaluation configuration: {"Skill Backbone":"GPT-5.4-mini","Frozen coding policy":"GPT-5.4-mini"}. This record does not assert independent reproduction. ### No-skill baseline Task Success Rate: 20. Result ID: 6a157da1fbc373592c0dc421. [Reporting paper]() Recorded evaluation configuration: {"Skill Backbone":"-","Frozen coding policy":"GPT-5.4-mini"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.