Model profile
Chatglm3 6B
Z.aideveloper
2023-10-27release date
#202 / 267overall rank
9eval lineages
Evidence summary
Chatglm3 6B has an estimated overall rank of #202; its 90% source-sensitivity interval is #127–#222. Published evidence spans 9 evals and 4 of 7 behavior components. Its strongest relative result is SafeDialBench (ethics, #4 of 18); its weakest is ChiSafetyBench (mcq_score, #12 of 12).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗zai-org/chatglm3-6b
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 |
|---|---|---|---|---|---|
| ChiSafetyBenchharmful_response_rate | #12 / 14 | 1.08 | ↓ lower | Source ↗official | |
| ChiSafetyBenchmcq_score | #12 / 12 | 41.16 | ↑ higher | Source ↗official | |
| CMoralEvalfamilial_morality | #11 / 26 | 0.43 | ↑ higher | Source ↗official | |
| CMoralEvalinternet_ethics | #12 / 26 | 0.4 | ↑ higher | Source ↗official | |
| CMoralEvalpersonal_morality | #11 / 26 | 0.42 | ↑ higher | Source ↗official | |
| CMoralEvalprofessional_ethics | #11 / 26 | 0.43 | ↑ higher | Source ↗official | |
| CMoralEvalsocial_morality | #11 / 26 | 0.43 | ↑ higher | Source ↗official | |
| Fake Alignment (FINE)multiple_choice_safe_decision_rate | #9 / 14 | 45.33 | ↑ higher | Source ↗official | |
| Fake Alignment (FINE)open_ended_safe_response_rate | #9 / 14 | 94.67 | ↑ higher | Source ↗official | |
| FLAMESdata_protection | #10 / 13 | 38.16 | ↑ higher | Source ↗official | |
| FLAMESfairness | #7 / 13 | 37.75 | ↑ higher | Source ↗official | |
| FLAMESlegality | #12 / 13 | 28.26 | ↑ higher | Source ↗official | |
| FLAMESmorality | #6 / 13 | 44.78 | ↑ higher | Source ↗official | |
| FLAMESsafety | #7 / 13 | 32.63 | ↑ higher | Source ↗official | |
| S-Evalbase_en_overall | #12 / 22 | 57.7 | ↑ higher | Source ↗official | |
| SafeDialBenchaggression | #15 / 18 | 7.017 | ↑ higher | Source ↗official | |
| SafeDialBenchethics | #4 / 18 | 7.637 | ↑ higher | Source ↗official | |
| SafeDialBenchfairness | #13 / 18 | 7.187 | ↑ higher | Source ↗official | |
| SafeDialBenchlegality | #6 / 18 | 7.983 | ↑ higher | Source ↗official | |
| SafeDialBenchmorality | #9 / 18 | 7.243 | ↑ higher | Source ↗official | |
| SafeDialBenchprivacy | #9 / 18 | 7.567 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #18 / 33 | 12.72 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #17 / 33 | 12.05 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #17 / 33 | 11.01 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #17 / 33 | 13.16 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #17 / 33 | 12.57 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #14 / 33 | 17.75 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #20 / 33 | 92.55 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #23 / 33 | 92.21 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_malicious_use | #20 / 33 | 91.15 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #22 / 33 | 91.38 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #19 / 33 | 88.73 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #19 / 33 | 86.96 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #25 / 33 | 17.78 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #21 / 33 | 20.56 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #22 / 33 | 19.55 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #22 / 33 | 20.95 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #24 / 33 | 17.81 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #22 / 33 | 22.78 | ↑ higher | Source ↗official | |
| SORRY-Benchavg | #34 / 51 | 0.36 | ↓ lower | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #9 / 31 | 68.97 | ↑ higher | Source ↗official | |
| SuperCLUE Safetyresponsible_ai | #17 / 31 | 56.36 | ↑ higher | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #28 / 31 | 65.96 | ↑ higher | Source ↗official |