Overall Performance on MMG2Skill-Bench Average
69.64Success RateMMG2Skill
Evaluation Results
| Method | Links | |
|---|---|---|
| MMG2SkillBackbone=Claude-Opus-4.6, Prompting Variant=MMG2Skill2026.06 | 69.64 | |
| MMG2SkillBackbone=GPT-5.5, Prompting Variant=MMG2Skill2026.06 | 67.41 | |
| MMG2SkillBackbone=Claude-Sonnet-4.6, Prompting Variant=MMG2Skill2026.06 | 64.92 | |
| MMG2SkillBackbone=Kimi-K2.6, Prompting Variant=MMG2Skill2026.06 | 58.81 | |
| Raw GuideBackbone=Claude-Opus-4.6, Prompting Variant=Raw Guide2026.06 | 56.86 | |
| VanillaBackbone=Claude-Opus-4.6, Prompting Variant=Vanilla2026.06 | 56.85 | |
| Raw GuideBackbone=GPT-5.5, Prompting Variant=Raw Guide2026.06 | 54.64 | |
| MMG2SkillBackbone=Gemini-3.1-Pro-Preview, Prompting Variant=MMG2Skill2026.06 | 54.17 | |
| VanillaBackbone=GPT-5.5, Prompting Variant=Vanilla2026.06 | 54.08 | |
| MMG2SkillBackbone=Qwen3.6-Plus, Prompting Variant=MMG2Skill2026.06 | 51.94 | |
| Raw GuideBackbone=Claude-Sonnet-4.6, Prompting Variant=Raw Guide2026.06 | 48.89 | |
| VanillaBackbone=Claude-Sonnet-4.6, Prompting Variant=Vanilla2026.06 | 48.81 | |
| VanillaBackbone=Kimi-K2.6, Prompting Variant=Vanilla2026.06 | 44.36 | |
| Raw GuideBackbone=Kimi-K2.6, Prompting Variant=Raw Guide2026.06 | 44.08 | |
| VanillaBackbone=Gemini-3.1-Pro-Preview, Prompting Variant=Vanilla2026.06 | 35.82 | |
| Raw GuideBackbone=Gemini-3.1-Pro-Preview, Prompting Variant=Raw Guide2026.06 | 31.94 | |
| VanillaBackbone=Qwen3.6-Plus, Prompting Variant=Vanilla2026.06 | 26.67 | |
| Raw GuideBackbone=Qwen3.6-Plus, Prompting Variant=Raw Guide2026.06 | 23.89 |