Model profile
Spark3.0
2023-10-24release date
2eval lineages
Evidence summary
Published evidence spans 2 evals and 5 of 7 behavior components. Its strongest relative result is SuperCLUE Safety (instruction_attack, #3 of 31); its weakest is SuperCLUE Safety (traditional_safety, #28 of 31).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Official model documentation ↗Family-level model document · iflytek · first party
- Release source ↗direct research preview identity date
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Better | Source |
|---|---|---|---|---|---|
| CRiskEvaldeception_willingness | #7 / 17 | 21.74 | ↓ lower | Source ↗official | |
| CRiskEvaldesire_for_resource | #13 / 17 | 38.82 | ↓ lower | Source ↗official | |
| CRiskEvalharmful_goal | #11 / 17 | 42.08 | ↓ lower | Source ↗official | |
| CRiskEvalimprovement_intent | #15 / 17 | 54.46 | ↓ lower | Source ↗official | |
| CRiskEvalmalicious_coordination | #10 / 17 | 15.56 | ↓ lower | Source ↗official | |
| CRiskEvalself_preservation | #9 / 17 | 37.12 | ↓ lower | Source ↗official | |
| CRiskEvalsituational_awareness | #11 / 17 | 66.02 | ↓ lower | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #3 / 31 | 72.41 | ↑ higher | Source ↗official | |
| SuperCLUE Safetyresponsible_ai | #11 / 31 | 65.45 | ↑ higher | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #28 / 31 | 65.96 | ↑ higher | Source ↗official |