Risk Decision-making on DeXposure-Bench frozen 2025 (test)
0.72Precisionpersist. rules
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| persist. rulesPred.=persist., Decision=rules2026.06 | 0.72 | 0.004 | 0.0076 | — | — | — | 0.514 | |
| DeXposure-FM rulesPred.=DeXposure-FM, Decision=rules2026.06 | 0.6 | 0.01 | 0.019 | — | — | — | 0.257 | |
| DeXposure-FM LLM + safety gatePred.=DeXposure-FM, Decision=LLM + safety gate2026.06 | 0.58 | 0.012 | 0.0234 | 2.45 | 0.437 | 1 | 0.435 | |
| LLM (no FM)Decision=LLM (no FM)2026.06 | 0.575 | 0.009 | 0.0184 | 2.24 | — | 1 | 0.532 | |
| DeXposure-FM LLMPred.=DeXposure-FM, Decision=LLM2026.06 | 0.57 | 0.012 | 0.0241 | 2.41 | 0.448 | 1 | 0.488 |