Instruction Following on IFEval (Score, Relative Gain)
86.5IFEval ScoreDeepSeek V3.2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepSeek V3.2SLM=DeepSeek V3.22026.05 | 86.5 | — | |
| +PACESLM=Ministral-3-14B2026.05 | 83.7 | 7.2 | |
| +CESLM=Ministral-3-14B2026.05 | 82.8 | 6 | |
| GEPASLM=Ministral-3-14B2026.05 | 81.6 | 4.5 | |
| +PESLM=Ministral-3-14B2026.05 | 81.6 | 4.5 | |
| Gödel AgentSLM=Ministral-3-14B2026.05 | 81 | 3.7 | |
| MIPROv2SLM=Ministral-3-14B2026.05 | 80.1 | 2.6 | |
| ACESLM=Ministral-3-14B2026.05 | 78.8 | 0.1 | |
| VanillaSLM=Ministral-3-14B2026.05 | 78.1 | — | |
| +PACESLM=Qwen3.5-9B2026.05 | 76.1 | 9.2 | |
| Gödel AgentSLM=Qwen3.5-9B2026.05 | 73.2 | 5 | |
| +PACESLM=Qwen3-4B2026.05 | 72.7 | 5.2 | |
| +CESLM=Qwen3.5-9B2026.05 | 72.5 | 4 | |
| +CESLM=Qwen3-4B2026.05 | 72.2 | 4.4 | |
| ACESLM=Qwen3.5-9B2026.05 | 71.8 | 3 | |
| +PESLM=Qwen3.5-9B2026.05 | 71.8 | 3 | |
| ACESLM=Qwen3-4B2026.05 | 71.1 | 2.9 | |
| Gödel AgentSLM=Qwen3-4B2026.05 | 70.5 | 2 | |
| +PESLM=Qwen3-4B2026.05 | 70.5 | 2 | |
| GEPASLM=Qwen3.5-9B2026.05 | 70.4 | 0.1 | |
| VanillaSLM=Qwen3.5-9B2026.05 | 69.7 | — | |
| MIPROv2SLM=Qwen3.5-9B2026.05 | 69.7 | 0 | |
| VanillaSLM=Qwen3-4B2026.05 | 69.1 | — | |
| MIPROv2SLM=Qwen3-4B2026.05 | 69.1 | 0 | |
| GEPASLM=Qwen3-4B2026.05 | 67.6 | -0.1 |