|
<!DOCTYPE html> |
|
<html lang="en"> |
|
<head> |
|
<meta charset="UTF-8"> |
|
<meta name="viewport" content="width=device-width, initial-scale=1.0"> |
|
<title>Guerra LLM Ranking</title> |
|
<script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.1/dist/chart.umd.min.js"></script> |
|
|
|
</head> |
|
<body> |
|
<canvas id="radarChart" height="750"></canvas> |
|
|
|
<script> |
|
let data = { |
|
labels: ['MMLU', 'MT-bench','Arena Elo'], |
|
datasets: [ |
|
{ |
|
label: 'GPT-4-Turbo', |
|
data: [null, 93.2,124.9], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'GPT-4-0314', |
|
data: [86.4, 86.4,119.0], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'GPT-3.5-Turbo-0314', |
|
data: [70.0, 79.4,112.3], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Mistral Medium', |
|
data: [75.3, 86.1,115.0], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Mixtral 8x7B Instruct v0.1', |
|
data: [70.6, 83.0,112.3], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Claude 2.0', |
|
data: [78.5, 80.6,113.1], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Claude 1.0', |
|
data: [77.0, 79.0,114.9], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Claude Instant 1', |
|
data: [73.4, 78.5,110.9], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Gemini Pro', |
|
data: [71.8, null,111.4], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Yi 34B Chat', |
|
data: [73.5, null,111.1], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'Falcon 180B Chat', |
|
data: [68.0, null,103.1], |
|
borderWidth: 2 |
|
}, |
|
{ |
|
label: 'LLama 2 70B Chat', |
|
data: [63.0, 68.6,107.9], |
|
borderWidth: 2 |
|
}, |
|
] |
|
}; |
|
|
|
let options = { |
|
responsive: true, |
|
maintainAspectRatio: false, |
|
scale: { |
|
ticks: { |
|
stepSize: 10, |
|
} |
|
}, |
|
}; |
|
|
|
let ctx = document.getElementById('radarChart').getContext('2d'); |
|
new Chart(ctx, { |
|
type: 'radar', |
|
data: data, |
|
options: options |
|
}); |
|
</script> |
|
</body> |
|
</html> |
