DeepData LLM Probe — Comparison
| Metric | deepseek/deepseek-chat | google/gemini-2.5-flash | meta-llama/llama-3.3-70b-instruct |
|---|---|---|---|
| Alignment gap (overall) | 0.285 | 0.332 | 0.014 |
| Neutral control | 0.224 | 0.216 | 0.022 |
| neutral-control | 0.224 | 0.216 | 0.022 |
| refusal-bait | 0.196 | 0.498 | 0.007 |
| risk | 0.374 | 0.166 | 0.022 |