Model profile
InternLM2 Chat 7B
Evidence summary
Safety. InternLM2 Chat 7B has an estimated Safety rank of #98; its 90% source-sensitivity interval is #32–#295. Its behavior-only rank is #101; company governance moves the combined estimate to #98. Published Safety evidence spans 4 eval lineages and 4 of 7 components. Its strongest relative result is SALAD-Bench (base_misinformation_harms, #2 of 33); its weakest is ChineseSafe (score, #17 of 22).
Freedom. InternLM2 Chat 7B has an estimated Freedom rank of #413; its 90% source-sensitivity interval is #270–#544. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is SALAD-Bench (base_information_safety_harms, #16 of 33); its weakest is SALAD-Bench (base_representation_toxicity, #31 of 33).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗internlm/internlm2-chat-7b
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗direct research preview identity date
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| ChineseSafescore | #17 / 22 | ↑49.49 | Source ↗official | |
| CMoralEvalfamilial_morality | #5 / 26 | ↑0.51 | Source ↗official | |
| CMoralEvalinternet_ethics | #4 / 26 | ↑0.51 | Source ↗official | |
| CMoralEvalpersonal_morality | #4 / 26 | ↑0.5 | Source ↗official | |
| CMoralEvalprofessional_ethics | #4 / 26 | ↑0.52 | Source ↗official | |
| CMoralEvalsocial_morality | #4 / 26 | ↑0.51 | Source ↗official | |
| JailBenchjailbreak_success_rate | #5 / 14 | ↓51.22 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #11 / 33 | ↑21.12 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #14 / 33 | ↑17.59 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #8 / 33 | ↑24.31 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #16 / 33 | ↑16.45 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #13 / 33 | ↑19.27 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #12 / 33 | ↑20.35 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #7 / 33 | ↑98.02 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #18 / 33 | ↑94.45 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #3 / 33 | ↑98.63 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #2 / 33 | ↑98.67 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #3 / 33 | ↑97.19 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #5 / 33 | ↑94.71 | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #5 / 33 | ↑63.89 | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #5 / 33 | ↑57.5 | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #5 / 33 | ↑62.24 | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #5 / 33 | ↑62.62 | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #5 / 33 | ↑61.25 | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #5 / 33 | ↑56.11 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| JailBenchjailbreak_success_rate | #10 / 14 | ↑51.22 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #22 / 33 | ↓21.12 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #19 / 33 | ↓17.59 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #26 / 33 | ↓24.31 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #18 / 33 | ↓16.45 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #21 / 33 | ↓19.27 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #22 / 33 | ↓20.35 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #27 / 33 | ↓98.02 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #16 / 33 | ↓94.45 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #31 / 33 | ↓98.63 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #31 / 33 | ↓98.67 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #31 / 33 | ↓97.19 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #29 / 33 | ↓94.71 | Source ↗official |