List your company
ResearchFormerly WizWand

Repository-level issue resolution on SWE-Bench Verified (held-out (sample of 150))

66Task Pass RateCODESKILL
45.896851.115956.33561.5541May 25, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=Qwen3.5-35B-A3B2026.056671.99
Prompt Skill Mgmt.Skill Backbone=GPT-5.4-mini, Frozen coding policy=Qwen3.5-35B-A3B2026.0564.6775.97
SFT-CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=Qwen3.5-35B-A3B2026.056478.65
Subtask MemorySkill Backbone=GPT-5.4-mini, Frozen coding policy=Qwen3.5-35B-A3B2026.0561.3381.31
Prompt Skill Mgmt.Skill Backbone=Qwen3.5-4B, Frozen coding policy=Qwen3.5-35B-A3B2026.0558.6776.69
No-skill baselineSkill Backbone=-, Frozen coding policy=Qwen3.5-35B-A3B2026.0557.3388.41
Prompt Skill Mgmt.Skill Backbone=GPT-5.4-mini, Frozen coding policy=GPT-5.4-mini2026.0556.676.53
CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=GPT-5.4-mini2026.05566.43
SFT-CODESKILLSkill Backbone=Qwen3.5-4B, Frozen coding policy=GPT-5.4-mini2026.0554.676.66
Subtask MemorySkill Backbone=GPT-5.4-mini, Frozen coding policy=GPT-5.4-mini2026.0552.676.73
Prompt Skill Mgmt.Skill Backbone=Qwen3.5-4B, Frozen coding policy=GPT-5.4-mini2026.0550.676.32
No-skill baselineSkill Backbone=-, Frozen coding policy=GPT-5.4-mini2026.0546.677.56