Meta-prompt-following sob pressão pra preservar propósito — uma capacidade que nenhum benchmark público avalia diretamente. Aqui: delta de score antes/depois em cada modelo do corpus.
| # | modelo | lab | antes | depois | Δ | latência | prompts |
|---|---|---|---|---|---|---|---|
| 1 | Jamba Large 1.7 (AI21) jamba-large-1.7 | AI21 Labs (Israel)trial Arquitetura Mamba-Transformer híbrida — única não-transformer pura do benchmark | 44.5 | 94.3 | +49.8 | 4.3s | 50 |
| 2 | Claude Sonnet (via CLI) claude-sonnet-4-6 (via CLI) | Anthropic (EUA)paid Claude Sonnet via CLI — versão mais rápida e leve que Opus | 44.5 | 93.6 | +49.0 | 18.6s | 50 |
| 3 | Claude (via CLI) claude-opus-4-6 (via CLI) | Anthropic (EUA)paid Claude Opus via CLI subscription — sem API key separada | 44.5 | 92.7 | +48.1 | 14.3s | 50 |
| 4 | Mistral Small mistral-small-latest | Mistral AI (França)free Modelo europeu open-weight focado em eficiência | 44.5 | 89.5 | +45.0 | 2.3s | 50 |
| 5 | Llama 3.3 70B (Groq) llama-3.3-70b-versatile | Meta (EUA) via Groqfree Llama open-source com inferência em hardware LPU da Groq | 44.5 | 89.4 | +44.8 | 3.3s | 50 |
| 6 | DeepSeek R1 deepseek-reasoner | DeepSeek (China)free Versão reasoner do DeepSeek com chain-of-thought explícito | 44.5 | 89.3 | +44.7 | 40.2s | 50 |
| 7 | DeepSeek V3 deepseek-chat | DeepSeek (China)free Modelo chinês de alta performance treinado com foco em código e raciocínio | 44.5 | 89.2 | +44.6 | 7.7s | 50 |
| 8 | Gemini 2.5 Flash gemini-2.5-flash | Google (EUA)free Modelo multimodal de referência do Google | 44.5 | 84.5 | +40.0 | 8.9s | 50 |
Cada provider aparece como coluna. Clique num provider pra abrir a lista expandida com o resultado em cada prompt do corpus — útil pra entender onde cada modelo afia melhor ou pior.
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| legal-advisor-en | ? | 0 | 100 | +100 |
| rigid-code-agent-en | ? | 1 | 100 | +99 |
| rigid-code-agent-pt | ? | 1 | 97 | +96 |
| agente-codigo-rigido | ? | 1 | 97 | +96 |
| legal-advisor-pt | ? | 0 | 93 | +93 |
| consultor-juridico | ? | 0 | 90 | +90 |
| medical-assistant-pt | ? | 0 | 87 | +87 |
| legal-advisor-es | ? | 14 | 100 | +86 |
| medical-assistant-en | ? | 0 | 86 | +86 |
| disguised-repetitions-pt | ? | 13 | 97 | +84 |
| repeticoes-disfarcadas | ? | 13 | 97 | +84 |
| assistente-medico | ? | 0 | 80 | +80 |
| rigid-code-agent-es | ? | 22 | 97 | +75 |
| disguised-repetitions-en | ? | 23 | 97 | +74 |
| marketing-strategist-pt | ? | 28 | 100 | +72 |
| estrategista-marketing | ? | 28 | 100 | +72 |
| medical-assistant-es | ? | 24 | 93 | +69 |
| disguised-repetitions-es | ? | 28 | 97 | +69 |
| marketing-strategist-en | ? | 43 | 100 | +57 |
| Supply chain optimizer (en) | en | 47 | 100 | +53 |
| financial-analyst-en | ? | 37 | 90 | +53 |
| Tax advisor (en) | en | 48 | 100 | +52 |
| Tutor de secundaria (es) | es | 49 | 100 | +51 |
| Consultor fiscal (es) | es | 51 | 100 | +49 |
| financial-analyst-pt | ? | 31 | 80 | +49 |
| Tutor de ensino médio (pt) | pt | 49 | 97 | +48 |
| tone-domain-mismatch-es | ? | 45 | 93 | +48 |
| Otimizador logístico (pt) | pt | 55 | 100 | +45 |
| High school tutor (en) | en | 53 | 97 | +44 |
| Consultor tributário (pt) | pt | 51 | 93 | +42 |
| task-style-no-roleplay-pt | ? | 52 | 90 | +38 |
| Optimizador logístico (es) | es | 63 | 100 | +37 |
| marketing-strategist-es | ? | 64 | 100 | +36 |
| task-style-no-roleplay-en | ? | 48 | 83 | +35 |
| financial-analyst-es | ? | 56 | 90 | +34 |
| task-style-no-roleplay-es | ? | 65 | 97 | +32 |
| contradiction-es | ? | 61 | 86 | +25 |
| unresolved-reference-es | ? | 79 | 100 | +21 |
| contradiction-en | ? | 61 | 79 | +18 |
| threat-framing-en | ? | 83 | 100 | +17 |
| unresolved-reference-pt | ? | 69 | 86 | +17 |
| threat-framing-es | ? | 79 | 93 | +14 |
| unresolved-reference-en | ? | 76 | 90 | +14 |
| role-inflation-pt | ? | 88 | 100 | +12 |
| contradiction-pt | ? | 61 | 72 | +11 |
| tone-domain-mismatch-pt | ? | 93 | 100 | +7 |
| tone-domain-mismatch-en | ? | 93 | 100 | +7 |
| role-inflation-en | ? | 94 | 100 | +6 |
| threat-framing-pt | ? | 90 | 93 | +3 |
| role-inflation-es | ? | 97 | 100 | +3 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| medical-assistant-pt | ? | 0 | 94 | +94 |
| legal-advisor-en | ? | 0 | 93 | +93 |
| consultor-juridico | ? | 0 | 90 | +90 |
| agente-codigo-rigido | ? | 1 | 90 | +89 |
| disguised-repetitions-pt | ? | 13 | 100 | +87 |
| repeticoes-disfarcadas | ? | 13 | 100 | +87 |
| legal-advisor-pt | ? | 0 | 86 | +86 |
| rigid-code-agent-en | ? | 1 | 87 | +86 |
| medical-assistant-en | ? | 0 | 83 | +83 |
| assistente-medico | ? | 0 | 83 | +83 |
| rigid-code-agent-pt | ? | 1 | 81 | +80 |
| disguised-repetitions-en | ? | 23 | 100 | +77 |
| disguised-repetitions-es | ? | 28 | 100 | +72 |
| legal-advisor-es | ? | 14 | 79 | +65 |
| rigid-code-agent-es | ? | 22 | 87 | +65 |
| estrategista-marketing | ? | 28 | 93 | +65 |
| medical-assistant-es | ? | 24 | 86 | +62 |
| marketing-strategist-pt | ? | 28 | 86 | +58 |
| marketing-strategist-en | ? | 43 | 97 | +54 |
| Supply chain optimizer (en) | en | 47 | 100 | +53 |
| financial-analyst-pt | ? | 31 | 83 | +52 |
| task-style-no-roleplay-en | ? | 48 | 100 | +52 |
| Tax advisor (en) | en | 48 | 100 | +52 |
| Tutor de ensino médio (pt) | pt | 49 | 100 | +51 |
| Tutor de secundaria (es) | es | 49 | 100 | +51 |
| Consultor fiscal (es) | es | 51 | 100 | +49 |
| tone-domain-mismatch-es | ? | 45 | 93 | +48 |
| High school tutor (en) | en | 53 | 100 | +47 |
| financial-analyst-en | ? | 37 | 83 | +46 |
| Consultor tributário (pt) | pt | 51 | 93 | +42 |
| task-style-no-roleplay-pt | ? | 52 | 93 | +41 |
| contradiction-en | ? | 61 | 100 | +39 |
| Otimizador logístico (pt) | pt | 55 | 93 | +38 |
| financial-analyst-es | ? | 56 | 93 | +37 |
| Optimizador logístico (es) | es | 63 | 100 | +37 |
| contradiction-pt | ? | 61 | 97 | +36 |
| contradiction-es | ? | 61 | 97 | +36 |
| marketing-strategist-es | ? | 64 | 97 | +33 |
| unresolved-reference-pt | ? | 69 | 100 | +31 |
| task-style-no-roleplay-es | ? | 65 | 93 | +28 |
| unresolved-reference-en | ? | 76 | 100 | +24 |
| unresolved-reference-es | ? | 79 | 100 | +21 |
| threat-framing-en | ? | 83 | 100 | +17 |
| threat-framing-es | ? | 79 | 93 | +14 |
| threat-framing-pt | ? | 90 | 100 | +10 |
| tone-domain-mismatch-en | ? | 93 | 100 | +7 |
| role-inflation-es | ? | 97 | 100 | +3 |
| role-inflation-pt | ? | 88 | 87 | +-1 |
| role-inflation-en | ? | 94 | 90 | +-4 |
| tone-domain-mismatch-pt | ? | 93 | 79 | +-14 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| rigid-code-agent-pt | ? | 1 | 97 | +96 |
| consultor-juridico | ? | 0 | 90 | +90 |
| legal-advisor-pt | ? | 0 | 90 | +90 |
| rigid-code-agent-en | ? | 1 | 90 | +89 |
| medical-assistant-en | ? | 0 | 87 | +87 |
| disguised-repetitions-pt | ? | 13 | 100 | +87 |
| assistente-medico | ? | 0 | 86 | +86 |
| legal-advisor-es | ? | 14 | 100 | +86 |
| repeticoes-disfarcadas | ? | 13 | 97 | +84 |
| legal-advisor-en | ? | 0 | 83 | +83 |
| medical-assistant-pt | ? | 0 | 83 | +83 |
| disguised-repetitions-en | ? | 23 | 100 | +77 |
| agente-codigo-rigido | ? | 1 | 77 | +76 |
| disguised-repetitions-es | ? | 28 | 100 | +72 |
| rigid-code-agent-es | ? | 22 | 87 | +65 |
| medical-assistant-es | ? | 24 | 86 | +62 |
| marketing-strategist-pt | ? | 28 | 86 | +58 |
| financial-analyst-pt | ? | 31 | 86 | +55 |
| task-style-no-roleplay-en | ? | 48 | 100 | +52 |
| Tax advisor (en) | en | 48 | 100 | +52 |
| Tutor de ensino médio (pt) | pt | 49 | 100 | +51 |
| estrategista-marketing | ? | 28 | 78 | +50 |
| financial-analyst-en | ? | 37 | 86 | +49 |
| task-style-no-roleplay-pt | ? | 52 | 100 | +48 |
| High school tutor (en) | en | 53 | 100 | +47 |
| Supply chain optimizer (en) | en | 47 | 93 | +46 |
| Consultor fiscal (es) | es | 51 | 97 | +46 |
| marketing-strategist-en | ? | 43 | 87 | +44 |
| Tutor de secundaria (es) | es | 49 | 93 | +44 |
| Consultor tributário (pt) | pt | 51 | 93 | +42 |
| contradiction-pt | ? | 61 | 100 | +39 |
| contradiction-en | ? | 61 | 100 | +39 |
| contradiction-es | ? | 61 | 100 | +39 |
| task-style-no-roleplay-es | ? | 65 | 100 | +35 |
| tone-domain-mismatch-es | ? | 45 | 79 | +34 |
| Otimizador logístico (pt) | pt | 55 | 86 | +31 |
| Optimizador logístico (es) | es | 63 | 93 | +30 |
| unresolved-reference-en | ? | 76 | 100 | +24 |
| marketing-strategist-es | ? | 64 | 86 | +22 |
| threat-framing-es | ? | 79 | 100 | +21 |
| unresolved-reference-es | ? | 79 | 100 | +21 |
| threat-framing-en | ? | 83 | 100 | +17 |
| unresolved-reference-pt | ? | 69 | 86 | +17 |
| financial-analyst-es | ? | 56 | 71 | +15 |
| role-inflation-pt | ? | 88 | 100 | +12 |
| threat-framing-pt | ? | 90 | 100 | +10 |
| role-inflation-en | ? | 94 | 100 | +6 |
| tone-domain-mismatch-en | ? | 93 | 97 | +4 |
| role-inflation-es | ? | 97 | 97 | +0 |
| tone-domain-mismatch-pt | ? | 93 | 86 | +-7 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| legal-advisor-en | ? | 0 | 97 | +97 |
| medical-assistant-en | ? | 0 | 93 | +93 |
| repeticoes-disfarcadas | ? | 13 | 100 | +87 |
| rigid-code-agent-en | ? | 1 | 84 | +83 |
| legal-advisor-pt | ? | 0 | 79 | +79 |
| disguised-repetitions-pt | ? | 13 | 90 | +77 |
| disguised-repetitions-en | ? | 23 | 100 | +77 |
| agente-codigo-rigido | ? | 1 | 77 | +76 |
| medical-assistant-pt | ? | 0 | 76 | +76 |
| rigid-code-agent-pt | ? | 1 | 77 | +76 |
| legal-advisor-es | ? | 14 | 86 | +72 |
| consultor-juridico | ? | 0 | 69 | +69 |
| disguised-repetitions-es | ? | 28 | 97 | +69 |
| estrategista-marketing | ? | 28 | 94 | +66 |
| assistente-medico | ? | 0 | 65 | +65 |
| marketing-strategist-pt | ? | 28 | 91 | +63 |
| financial-analyst-pt | ? | 31 | 90 | +59 |
| rigid-code-agent-es | ? | 22 | 80 | +58 |
| medical-assistant-es | ? | 24 | 79 | +55 |
| Tutor de secundaria (es) | es | 49 | 100 | +51 |
| Supply chain optimizer (en) | en | 47 | 97 | +50 |
| tone-domain-mismatch-es | ? | 45 | 93 | +48 |
| task-style-no-roleplay-pt | ? | 52 | 100 | +48 |
| High school tutor (en) | en | 53 | 100 | +47 |
| financial-analyst-en | ? | 37 | 83 | +46 |
| task-style-no-roleplay-en | ? | 48 | 93 | +45 |
| Tax advisor (en) | en | 48 | 93 | +45 |
| Tutor de ensino médio (pt) | pt | 49 | 93 | +44 |
| marketing-strategist-en | ? | 43 | 86 | +43 |
| Consultor tributário (pt) | pt | 51 | 90 | +39 |
| Optimizador logístico (es) | es | 63 | 100 | +37 |
| task-style-no-roleplay-es | ? | 65 | 100 | +35 |
| Consultor fiscal (es) | es | 51 | 86 | +35 |
| marketing-strategist-es | ? | 64 | 94 | +30 |
| contradiction-es | ? | 61 | 90 | +29 |
| Otimizador logístico (pt) | pt | 55 | 83 | +28 |
| financial-analyst-es | ? | 56 | 83 | +27 |
| contradiction-pt | ? | 61 | 83 | +22 |
| threat-framing-es | ? | 79 | 100 | +21 |
| unresolved-reference-en | ? | 76 | 93 | +17 |
| unresolved-reference-es | ? | 79 | 93 | +14 |
| role-inflation-pt | ? | 88 | 100 | +12 |
| contradiction-en | ? | 61 | 73 | +12 |
| unresolved-reference-pt | ? | 69 | 79 | +10 |
| tone-domain-mismatch-pt | ? | 93 | 100 | +7 |
| tone-domain-mismatch-en | ? | 93 | 100 | +7 |
| role-inflation-en | ? | 94 | 100 | +6 |
| threat-framing-en | ? | 83 | 87 | +4 |
| role-inflation-es | ? | 97 | 100 | +3 |
| threat-framing-pt | ? | 90 | 79 | +-11 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| legal-advisor-en | ? | 0 | 91 | +91 |
| medical-assistant-en | ? | 0 | 90 | +90 |
| repeticoes-disfarcadas | ? | 13 | 100 | +87 |
| rigid-code-agent-en | ? | 1 | 87 | +86 |
| disguised-repetitions-pt | ? | 13 | 93 | +80 |
| consultor-juridico | ? | 0 | 79 | +79 |
| assistente-medico | ? | 0 | 79 | +79 |
| legal-advisor-pt | ? | 0 | 79 | +79 |
| medical-assistant-pt | ? | 0 | 79 | +79 |
| disguised-repetitions-en | ? | 23 | 97 | +74 |
| estrategista-marketing | ? | 28 | 93 | +65 |
| rigid-code-agent-pt | ? | 1 | 65 | +64 |
| financial-analyst-en | ? | 37 | 100 | +63 |
| medical-assistant-es | ? | 24 | 86 | +62 |
| marketing-strategist-pt | ? | 28 | 90 | +62 |
| financial-analyst-pt | ? | 31 | 93 | +62 |
| legal-advisor-es | ? | 14 | 72 | +58 |
| disguised-repetitions-es | ? | 28 | 86 | +58 |
| agente-codigo-rigido | ? | 1 | 58 | +57 |
| rigid-code-agent-es | ? | 22 | 79 | +57 |
| marketing-strategist-en | ? | 43 | 97 | +54 |
| Supply chain optimizer (en) | en | 47 | 100 | +53 |
| task-style-no-roleplay-en | ? | 48 | 100 | +52 |
| Tax advisor (en) | en | 48 | 100 | +52 |
| Consultor fiscal (es) | es | 51 | 100 | +49 |
| High school tutor (en) | en | 53 | 97 | +44 |
| Consultor tributário (pt) | pt | 51 | 93 | +42 |
| tone-domain-mismatch-es | ? | 45 | 86 | +41 |
| task-style-no-roleplay-pt | ? | 52 | 93 | +41 |
| contradiction-en | ? | 61 | 100 | +39 |
| Tutor de ensino médio (pt) | pt | 49 | 86 | +37 |
| Tutor de secundaria (es) | es | 49 | 86 | +37 |
| task-style-no-roleplay-es | ? | 65 | 100 | +35 |
| contradiction-es | ? | 61 | 93 | +32 |
| Otimizador logístico (pt) | pt | 55 | 83 | +28 |
| contradiction-pt | ? | 61 | 86 | +25 |
| Optimizador logístico (es) | es | 63 | 86 | +23 |
| unresolved-reference-es | ? | 79 | 100 | +21 |
| financial-analyst-es | ? | 56 | 76 | +20 |
| marketing-strategist-es | ? | 64 | 80 | +16 |
| unresolved-reference-en | ? | 76 | 90 | +14 |
| role-inflation-pt | ? | 88 | 100 | +12 |
| threat-framing-en | ? | 83 | 90 | +7 |
| threat-framing-es | ? | 79 | 86 | +7 |
| tone-domain-mismatch-pt | ? | 93 | 100 | +7 |
| tone-domain-mismatch-en | ? | 93 | 100 | +7 |
| role-inflation-en | ? | 94 | 100 | +6 |
| threat-framing-pt | ? | 90 | 93 | +3 |
| role-inflation-es | ? | 97 | 100 | +3 |
| unresolved-reference-pt | ? | 69 | 72 | +3 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| rigid-code-agent-en | ? | 1 | 97 | +96 |
| legal-advisor-en | ? | 0 | 93 | +93 |
| legal-advisor-pt | ? | 0 | 86 | +86 |
| agente-codigo-rigido | ? | 1 | 86 | +85 |
| repeticoes-disfarcadas | ? | 13 | 97 | +84 |
| consultor-juridico | ? | 0 | 83 | +83 |
| disguised-repetitions-pt | ? | 13 | 94 | +81 |
| assistente-medico | ? | 0 | 79 | +79 |
| medical-assistant-en | ? | 0 | 77 | +77 |
| legal-advisor-es | ? | 14 | 90 | +76 |
| medical-assistant-pt | ? | 0 | 76 | +76 |
| disguised-repetitions-en | ? | 23 | 97 | +74 |
| rigid-code-agent-pt | ? | 1 | 74 | +73 |
| disguised-repetitions-es | ? | 28 | 100 | +72 |
| estrategista-marketing | ? | 28 | 93 | +65 |
| financial-analyst-pt | ? | 31 | 94 | +63 |
| rigid-code-agent-es | ? | 22 | 84 | +62 |
| medical-assistant-es | ? | 24 | 83 | +59 |
| Supply chain optimizer (en) | en | 47 | 100 | +53 |
| task-style-no-roleplay-en | ? | 48 | 100 | +52 |
| financial-analyst-en | ? | 37 | 87 | +50 |
| Tax advisor (en) | en | 48 | 97 | +49 |
| tone-domain-mismatch-es | ? | 45 | 93 | +48 |
| Tutor de secundaria (es) | es | 49 | 97 | +48 |
| High school tutor (en) | en | 53 | 100 | +47 |
| marketing-strategist-en | ? | 43 | 87 | +44 |
| Consultor tributário (pt) | pt | 51 | 93 | +42 |
| Consultor fiscal (es) | es | 51 | 93 | +42 |
| task-style-no-roleplay-pt | ? | 52 | 93 | +41 |
| marketing-strategist-pt | ? | 28 | 66 | +38 |
| financial-analyst-es | ? | 56 | 94 | +38 |
| Tutor de ensino médio (pt) | pt | 49 | 86 | +37 |
| Otimizador logístico (pt) | pt | 55 | 86 | +31 |
| Optimizador logístico (es) | es | 63 | 93 | +30 |
| marketing-strategist-es | ? | 64 | 93 | +29 |
| contradiction-pt | ? | 61 | 86 | +25 |
| contradiction-en | ? | 61 | 86 | +25 |
| contradiction-es | ? | 61 | 86 | +25 |
| task-style-no-roleplay-es | ? | 65 | 87 | +22 |
| unresolved-reference-en | ? | 76 | 93 | +17 |
| threat-framing-es | ? | 79 | 93 | +14 |
| threat-framing-en | ? | 83 | 93 | +10 |
| tone-domain-mismatch-en | ? | 93 | 100 | +7 |
| role-inflation-en | ? | 94 | 100 | +6 |
| role-inflation-pt | ? | 88 | 93 | +5 |
| unresolved-reference-pt | ? | 69 | 72 | +3 |
| unresolved-reference-es | ? | 79 | 79 | +0 |
| role-inflation-es | ? | 97 | 93 | +-4 |
| threat-framing-pt | ? | 90 | 83 | +-7 |
| tone-domain-mismatch-pt | ? | 93 | 79 | +-14 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| legal-advisor-en | ? | 0 | 97 | +97 |
| agente-codigo-rigido | ? | 1 | 90 | +89 |
| disguised-repetitions-pt | ? | 13 | 100 | +87 |
| repeticoes-disfarcadas | ? | 13 | 100 | +87 |
| legal-advisor-pt | ? | 0 | 86 | +86 |
| rigid-code-agent-en | ? | 1 | 87 | +86 |
| consultor-juridico | ? | 0 | 83 | +83 |
| rigid-code-agent-pt | ? | 1 | 83 | +82 |
| medical-assistant-pt | ? | 0 | 79 | +79 |
| medical-assistant-en | ? | 0 | 79 | +79 |
| assistente-medico | ? | 0 | 79 | +79 |
| rigid-code-agent-es | ? | 22 | 100 | +78 |
| disguised-repetitions-en | ? | 23 | 100 | +77 |
| disguised-repetitions-es | ? | 28 | 100 | +72 |
| legal-advisor-es | ? | 14 | 79 | +65 |
| medical-assistant-es | ? | 24 | 86 | +62 |
| financial-analyst-pt | ? | 31 | 93 | +62 |
| task-style-no-roleplay-en | ? | 48 | 100 | +52 |
| Tax advisor (en) | en | 48 | 100 | +52 |
| Tutor de ensino médio (pt) | pt | 49 | 100 | +51 |
| Tutor de secundaria (es) | es | 49 | 100 | +51 |
| Supply chain optimizer (en) | en | 47 | 97 | +50 |
| financial-analyst-en | ? | 37 | 86 | +49 |
| Consultor tributário (pt) | pt | 51 | 100 | +49 |
| task-style-no-roleplay-pt | ? | 52 | 100 | +48 |
| estrategista-marketing | ? | 28 | 76 | +48 |
| High school tutor (en) | en | 53 | 100 | +47 |
| marketing-strategist-pt | ? | 28 | 72 | +44 |
| Consultor fiscal (es) | es | 51 | 93 | +42 |
| tone-domain-mismatch-es | ? | 45 | 86 | +41 |
| Otimizador logístico (pt) | pt | 55 | 93 | +38 |
| Optimizador logístico (es) | es | 63 | 100 | +37 |
| task-style-no-roleplay-es | ? | 65 | 100 | +35 |
| marketing-strategist-es | ? | 64 | 93 | +29 |
| contradiction-es | ? | 61 | 86 | +25 |
| marketing-strategist-en | ? | 43 | 65 | +22 |
| contradiction-pt | ? | 61 | 79 | +18 |
| financial-analyst-es | ? | 56 | 73 | +17 |
| contradiction-en | ? | 61 | 77 | +16 |
| threat-framing-es | ? | 79 | 93 | +14 |
| unresolved-reference-en | ? | 76 | 86 | +10 |
| threat-framing-en | ? | 83 | 90 | +7 |
| unresolved-reference-es | ? | 79 | 86 | +7 |
| tone-domain-mismatch-en | ? | 93 | 97 | +4 |
| threat-framing-pt | ? | 90 | 93 | +3 |
| role-inflation-en | ? | 94 | 97 | +3 |
| role-inflation-pt | ? | 88 | 86 | +-2 |
| role-inflation-es | ? | 97 | 90 | +-7 |
| tone-domain-mismatch-pt | ? | 93 | 86 | +-7 |
| unresolved-reference-pt | ? | 69 | 58 | +-11 |
| prompt | lang | antes | depois | Δ |
|---|---|---|---|---|
| consultor-juridico | ? | 0 | 90 | +90 |
| agente-codigo-rigido | ? | 1 | 87 | +86 |
| rigid-code-agent-pt | ? | 1 | 86 | +85 |
| repeticoes-disfarcadas | ? | 13 | 97 | +84 |
| disguised-repetitions-pt | ? | 13 | 93 | +80 |
| legal-advisor-en | ? | 0 | 79 | +79 |
| disguised-repetitions-en | ? | 23 | 100 | +77 |
| rigid-code-agent-en | ? | 1 | 75 | +74 |
| medical-assistant-pt | ? | 0 | 72 | +72 |
| assistente-medico | ? | 0 | 69 | +69 |
| estrategista-marketing | ? | 28 | 97 | +69 |
| disguised-repetitions-es | ? | 28 | 93 | +65 |
| rigid-code-agent-es | ? | 22 | 86 | +64 |
| legal-advisor-pt | ? | 0 | 58 | +58 |
| legal-advisor-es | ? | 14 | 72 | +58 |
| medical-assistant-en | ? | 0 | 58 | +58 |
| medical-assistant-es | ? | 24 | 79 | +55 |
| tone-domain-mismatch-es | ? | 45 | 93 | +48 |
| High school tutor (en) | en | 53 | 100 | +47 |
| Supply chain optimizer (en) | en | 47 | 93 | +46 |
| task-style-no-roleplay-en | ? | 48 | 93 | +45 |
| Otimizador logístico (pt) | pt | 55 | 100 | +45 |
| Tutor de ensino médio (pt) | pt | 49 | 93 | +44 |
| Consultor tributário (pt) | pt | 51 | 93 | +42 |
| Tax advisor (en) | en | 48 | 90 | +42 |
| task-style-no-roleplay-pt | ? | 52 | 93 | +41 |
| marketing-strategist-en | ? | 43 | 81 | +38 |
| Consultor fiscal (es) | es | 51 | 86 | +35 |
| Tutor de secundaria (es) | es | 49 | 83 | +34 |
| financial-analyst-pt | ? | 31 | 63 | +32 |
| financial-analyst-en | ? | 37 | 68 | +31 |
| Optimizador logístico (es) | es | 63 | 93 | +30 |
| marketing-strategist-pt | ? | 28 | 55 | +27 |
| contradiction-pt | ? | 61 | 86 | +25 |
| contradiction-en | ? | 61 | 86 | +25 |
| task-style-no-roleplay-es | ? | 65 | 86 | +21 |
| contradiction-es | ? | 61 | 79 | +18 |
| financial-analyst-es | ? | 56 | 71 | +15 |
| threat-framing-en | ? | 83 | 97 | +14 |
| threat-framing-es | ? | 79 | 93 | +14 |
| threat-framing-pt | ? | 90 | 100 | +10 |
| unresolved-reference-en | ? | 76 | 86 | +10 |
| marketing-strategist-es | ? | 64 | 72 | +8 |
| unresolved-reference-es | ? | 79 | 86 | +7 |
| role-inflation-en | ? | 94 | 100 | +6 |
| role-inflation-es | ? | 97 | 100 | +3 |
| tone-domain-mismatch-en | ? | 93 | 93 | +0 |
| role-inflation-pt | ? | 88 | 83 | +-5 |
| tone-domain-mismatch-pt | ? | 93 | 83 | +-10 |
| unresolved-reference-pt | ? | 69 | 58 | +-11 |