Instruction Following on Instruction Following (test)
0.9NMSEMENTAT (Detailed Prompt)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MENTAT (Detailed Prompt)LM=GPT-OSS-20B2025.08 | 0.9 | 0.43 | |
| MENTAT (Basic Prompt)LM=GPT-OSS-20B2025.08 | 0.95 | 0.42 | |
| MENTAT (Basic Prompt)-AvgLM=GPT-OSS-20B2025.08 | 1.06 | 0.38 | |
| GEPALM=GPT-OSS-20B2025.08 | 1.06 | 0.46 | |
| Gradient DescentLM=NeoBERT2025.08 | 1.08 | 0.36 | |
| MENTAT (Detailed Prompt)-AvgLM=GPT-OSS-20B2025.08 | 1.09 | 0.39 | |
| Detailed PromptLM=GPT-OSS-20B2025.08 | 1.16 | 0.33 | |
| Basic PromptLM=GPT-OSS-20B2025.08 | 1.18 | 0.32 | |
| MENTAT (Detailed Prompt) PromptLM=GPT-OSS-20B2025.08 | 1.24 | 0.36 | |
| MENTAT (Basic Prompt) PromptLM=GPT-OSS-20B2025.08 | 1.25 | 0.35 | |
| RL Fine-TuningLM=GPT-OSS-20B2025.08 | 1.51 | 0.37 |