Model profile
InternLM Chat 7B
Evidence summary
Safety. InternLM Chat 7B has an estimated Safety rank of #209; its 90% source-sensitivity interval is #63–#279. Its behavior-only rank is #213; company governance moves the combined estimate to #209. Published Safety evidence spans 5 eval lineages and 4 of 7 components. Its strongest relative result is FLAMES (legality, #1 of 13); its weakest is SALAD-Bench (mcq_representation_toxicity, #30 of 33).
Freedom. InternLM Chat 7B has an estimated Freedom rank of #308; its 90% source-sensitivity interval is #76–#537. Published Freedom evidence spans 4 eval lineages and 1 of 1 components. Its strongest relative result is SuperCLUE Safety (traditional_safety, #2 of 31); its weakest is FLAMES (legality, #13 of 13).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗internlm/internlm-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 |
|---|---|---|---|---|
| Fake Alignment (FINE)multiple_choice_safe_decision_rate | #6 / 14 | ↑57.33 | Source ↗official | |
| Fake Alignment (FINE)open_ended_safe_response_rate | #11 / 14 | ↑92 | Source ↗official | |
| FLAMESdata_protection | #2 / 13 | ↑61.84 | Source ↗official | |
| FLAMESfairness | #2 / 13 | ↑44.58 | Source ↗official | |
| FLAMESlegality | #1 / 13 | ↑76.09 | Source ↗official | |
| FLAMESmorality | #3 / 13 | ↑51.24 | Source ↗official | |
| FLAMESsafety | #5 / 13 | ↑35.9 | Source ↗official | |
| SafetyBenchEM | #6 / 21 | ↑75.4 | Source ↗official | |
| SafetyBenchIA | #8 / 21 | ↑79.5 | Source ↗official | |
| SafetyBenchMH | #6 / 21 | ↑84.3 | Source ↗official | |
| SafetyBenchOFF | #12 / 21 | ↑67.2 | Source ↗official | |
| SafetyBenchPH | #7 / 21 | ↑74.15 | Source ↗official | |
| SafetyBenchPP | #7 / 21 | ↑78.7 | Source ↗official | |
| SafetyBenchUB | #8 / 21 | ↑64.75 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #11 / 33 | ↑21.12 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #16 / 33 | ↑16.61 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #12 / 33 | ↑22.51 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #11 / 33 | ↑19.9 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #9 / 33 | ↑23.71 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #11 / 33 | ↑22.08 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #10 / 33 | ↑96.85 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #15 / 33 | ↑95.13 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #12 / 33 | ↑96.28 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #5 / 33 | ↑97.05 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #7 / 33 | ↑94.37 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #15 / 33 | ↑90.95 | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #30 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #30 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #30 / 33 | ↑0.0641 | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #30 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #30 / 33 | ↑0.1042 | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #30 / 33 | ↑0 | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #20 / 31 | ↑58.62 | Source ↗official | |
| SuperCLUE Safetyresponsible_ai | #28 / 31 | ↑45.45 | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #28 / 31 | ↑65.96 | 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 |
|---|---|---|---|---|
| Fake Alignment (FINE)open_ended_safe_response_rate | #4 / 14 | ↓92 | Source ↗official | |
| FLAMESdata_protection | #12 / 13 | ↓61.84 | Source ↗official | |
| FLAMESfairness | #12 / 13 | ↓44.58 | Source ↗official | |
| FLAMESlegality | #13 / 13 | ↓76.09 | Source ↗official | |
| FLAMESmorality | #11 / 13 | ↓51.24 | Source ↗official | |
| FLAMESsafety | #9 / 13 | ↓35.9 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #22 / 33 | ↓21.12 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #18 / 33 | ↓16.61 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #22 / 33 | ↓22.51 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #23 / 33 | ↓19.9 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #25 / 33 | ↓23.71 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #23 / 33 | ↓22.08 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #24 / 33 | ↓96.85 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #19 / 33 | ↓95.13 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #22 / 33 | ↓96.28 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #29 / 33 | ↓97.05 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #27 / 33 | ↓94.37 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #19 / 33 | ↓90.95 | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #9 / 31 | ↓58.62 | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #2 / 31 | ↓65.96 | Source ↗official |