Harm Evaluation on PrincipalBench 36-item core
0Harm RateGemini-3p1-flash-lite
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-3p1-flash-liteInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 0 | — | — | |
| Mistral-LargeInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 6 | — | — | |
| Mistral-LargeInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 9 | — | — | |
| DeepSeek-v3.1Interaction Arm=prompted, Evaluation Seeds=seed-12026.06 | 9 | — | — | |
| DeepSeek-v3.1Interaction Arm=plain, Evaluation Seeds=seed-12026.06 | 11 | — | — | |
| DeepSeek-v3.1Interaction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 11 | — | — | |
| Claude-OpusInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 11 | — | — | |
| Llama-3.1-70BInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 11 | — | — | |
| Qwen3-32BInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 12 | — | — | |
| Gemini-2.5-flashInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 14 | — | — | |
| Qwen3-32BInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 14 | — | — | |
| Gemini-3-flashInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 15 | — | — | |
| Gemini-2.5-flashInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 17 | — | — | |
| Gemini-3p1-flash-liteInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 17 | — | — | |
| Gemini-3p1-flash-liteInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 17 | — | — | |
| Llama-3.1-70BInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 17 | — | — | |
| Gemini-3-flashInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 17 | — | — | |
| Claude-SonnetInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 17 | — | — | |
| Gemini-2.5-flashInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 19 | — | — | |
| Claude-OpusInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 19 | — | — | |
| Claude-OpusInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 19 | — | — | |
| Claude-SonnetInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 19 | — | — | |
| Gemini-3-flashInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 20 | — | — | |
| Llama-3.1-70BInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 21 | — | — | |
| Mistral-LargeInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 22 | — | — | |
| Claude-SonnetInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 22 | — | — | |
| Qwen3-32BInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 25 | — | — | |
| GLM-4.6Interaction Arm=plain, Evaluation Seeds=seed-12026.06 | 43 | — | — | |
| GLM-4.6Interaction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 43 | — | — | |
| GPT-5-miniInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 44 | — | — | |
| Qwen3.5-27BInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 59 | — | — | |
| GLM-4.6Interaction Arm=prompted, Evaluation Seeds=seed-12026.06 | 61 | — | — | |
| GPT-5Interaction Arm=plain, Evaluation Seeds=seed-12026.06 | 63 | — | — | |
| GPT-5-miniInteraction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 65 | — | — | |
| GPT-5Interaction Arm=prompted, Evaluation Seeds=seed-12026.06 | 71 | — | — | |
| GPT-5Interaction Arm=scaffolded, Evaluation Seeds=seed-12026.06 | 71 | — | — | |
| Qwen3.5-27BInteraction Arm=plain, Evaluation Seeds=seed-12026.06 | 72 | — | — | |
| GPT-5-miniInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 76 | — | — | |
| Qwen3.5-27BInteraction Arm=prompted, Evaluation Seeds=seed-12026.06 | 79 | — | — | |
| Claude-OpusEvaluation Seeds=n=52026.06 | — | 18.1 | 2.8 | |
| Claude-SonnetEvaluation Seeds=n=52026.06 | — | 19.5 | 1.5 | |
| DeepSeek-v3.1Evaluation Seeds=n=52026.06 | — | 12.3 | 2.5 | |
| Gemini-2.5-flashEvaluation Seeds=n=52026.06 | — | 5.5 | 6.3 | |
| Gemini-3-flashEvaluation Seeds=n=52026.06 | — | 19.4 | 2.3 | |
| Gemini-3p1-flash-liteEvaluation Seeds=n=52026.06 | — | 12 | 2.1 | |
| GLM-4.6Evaluation Seeds=n=52026.06 | — | 46 | 2.9 | |
| GPT-5Evaluation Seeds=n=52026.06 | — | 71.1 | 2.2 | |
| GPT-5-miniEvaluation Seeds=n=52026.06 | — | 53.6 | 5.1 | |
| Llama-3.1-70BEvaluation Seeds=n=52026.06 | — | 19.2 | 2.7 | |
| Mistral-LargeEvaluation Seeds=n=52026.06 | — | 11 | 2.8 | |
| Qwen3-32BEvaluation Seeds=n=52026.06 | — | 16.5 | 2.6 | |
| Qwen3.5-27BEvaluation Seeds=n=52026.06 | — | 75.3 | 2.9 |