Trustworthiness Evaluation on LLM Trustworthiness Benchmark
89.58Bias ScoreMi:dm K 2.5 Pro (March ‘26)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Mi:dm K 2.5 Pro (March ‘26)Reasoning=On2026.03 | 89.58 | 85.67 | 97.5 | 87.22 | 88.33 | |
| Qwen3-30B-A3BReasoning=On2026.03 | 85.49 | 83.75 | 95.83 | 80 | 84.5 | |
| Mi:dm K 2.5 Pro (March ‘26)Reasoning=Off2026.03 | 84.79 | 76.58 | 98.33 | 82.22 | 82.44 | |
| EXAONE 3.5number of parameters=32B2025.09 | 84.5 | 84 | 96.3 | 91.2 | 87.5 | |
| HyperCLOVAX-SEED-Think-32BReasoning=On2026.03 | 84.03 | 80.25 | 97.08 | 83.34 | 83.5 | |
| Mi:dm 2.0 Base-instVersion=Base, Mode=Instruction-tuned2026.01 | 80.77 | 81.45 | 95.83 | 82.74 | 83.61 | |
| Mi:dm 2.0-Base2025.09 | 80.7 | 81.4 | 95.8 | 82.7 | 83.6 | |
| Mi:dm 2.0 Mini-instVersion=Mini, Mode=Instruction-tuned2026.01 | 79.15 | 77.71 | 85 | 72.1 | 78.44 | |
| Qwen3-30B-A3BReasoning=Off2026.03 | 77.29 | 73.08 | 95.83 | 74.44 | 76.55 | |
| Exaone-3.5-7.8B-instParameters=7.8B, Mode=Instruction-tuned2026.01 | 75.5 | 71.86 | 93.75 | 81.56 | 77.71 | |
| EXAONE 3.5number of parameters=7.8B2025.09 | 75.5 | 71.8 | 93.7 | 81.5 | 77.7 | |
| Llama-3.1-8B-instParameters=8B, Mode=Instruction-tuned2026.01 | 72.78 | 70 | 87.08 | 73.47 | 72.94 | |
| Llama-3.1number of parameters=8B2025.09 | 72.7 | 70 | 87 | 73.47 | 72.9 | |
| EXAONE-4.0-32BReasoning=Off2026.03 | 70.21 | 73.41 | 95.83 | 74.3 | 73.8 | |
| HyperCLOVAX-SEED-Think-32BReasoning=Off2026.03 | 69.79 | 65.83 | 95 | 66.94 | 69.58 | |
| Exaone-3.5-2.4B-instParameters=2.4B, Mode=Instruction-tuned2026.01 | 64.84 | 60.8 | 76.25 | 68.74 | 66.53 | |
| EXAONE-4.0-32BReasoning=On2026.03 | 63.13 | 65.08 | 75.42 | 62.64 | 64.5 |