Dense vs MoE Neural Network Infographic
A technical infographic comparing Dense and Mixture of Experts (MoE) AI models with network diagrams and bullet points.
Prompt
{
"type": "資訊圖表比較圖表",
"header": {
"title": "{argument name=\"主標題\" default=\"Dense 與 MoE 的差異\"}"
},
"layout": {
"structure": "兩個主要欄位由中央 VS 徽章分隔，底部有分割頁尾",
"sections": [
{
"position": "左欄",
"theme_color": "blue",
"header": "{argument name=\"左模型名稱\" default=\"Dense 模型\"}",
"subtitle": "所有神經元都被啟動",
"diagram": {
"type": "完全連接神經網路",
"elements": [
"1 個標記為輸入的橙色輸入節點",
"4 個隱藏層，分別有 4、5、4 和 2 個節點",
"節點顏色為白色、藍色和黃色",
"所有相鄰節點之間有密集的交錯連線"
]
},
"bullet_points": {
"count": 2,
"items": [
"{argument name=\"左項目符號\" default=\"使用所有參數\"}",
"計算成本高"
]
}
},
{
"position": "右欄",
"theme_color": "orange",
"header": "{argument name=\"右模型名稱\" default=\"MoE 模型\"}",
"subtitle": "部分專家被選擇性啟動",
"diagram": {
"type": "專家混合網路",
"elements": [
"1 個標記為輸入的橙色輸入節點",
"3 個矩形區塊，標記為 Expert 1, Expert 2, Expert 3",
"1 個標記為輸出的黃色輸出節點",
"分支箭頭連接輸入到專家，以及專家到輸出"
]
},
"bullet_points": {
"count": 2,
"items": [
"{argument name=\"右項目符號\" default=\"僅使用部分專家\"}",
"高效且可擴展"
]
}
},
{
"position": "center",
"element": "帶有 VS 文字的紅色圓形徽章",
"connections": "藍色箭頭指向左，橙色箭頭指向右"
},
{
"position": "頁尾左",
"background": "light blue",
"text": "Dense: 所有層常時運行，使用所有參數",
"icon": "1 個 CPU 晶片圖形",
"label": "高耗電量"
},
{
"position": "頁尾右",
"background": "light orange",
"text": "MoE: 僅動員必要專家",
"icons": "2 個圓形圖形（橙色箭頭、藍色閃電符號）",
"label": "低成本・高效率"
}
]
}
}
機器翻譯 — 實際產出作品的是原文 prompt。
原文 Prompt
{
"type": "infographic comparison diagram",
"header": {
"title": "{argument name=\"main title\" default=\"Dense と MoE の違い\"}"
},
"layout": {
"structure": "two main columns separated by a central VS badge, with a split footer at the bottom",
"sections": [
{
"position": "left column",
"theme_color": "blue",
"header": "{argument name=\"left model name\" default=\"Dense モデル\"}",
"subtitle": "全てのニューロンが活性化",
"diagram": {
"type": "fully connected neural network",
"elements": [
"1 orange input node labeled 入力",
"4 hidden layers with 4, 5, 4, and 2 nodes respectively",
"nodes colored white, blue, and yellow",
"dense intersecting connection lines between all adjacent nodes"
]
},
"bullet_points": {
"count": 2,
"items": [
"{argument name=\"left bullet point\" default=\"全てのパラメータが使用\"}",
"計算コストが高い"
]
}
},
{
"position": "right column",
"theme_color": "orange",
"header": "{argument name=\"right model name\" default=\"MoE モデル\"}",
"subtitle": "一部の専門家が選択的に活性化",
"diagram": {
"type": "mixture of experts network",
"elements": [
"1 orange input node labeled 入力",
"3 rectangular blocks labeled Expert 1, Expert 2, Expert 3",
"1 yellow output node labeled 出力",
"branching arrows connecting input to experts, and experts to output"
]
},
"bullet_points": {
"count": 2,
"items": [
"{argument name=\"right bullet point\" default=\"一部のエキスパートのみ使用\"}",
"効率的でスケーラブル"
]
}
},
{
"position": "center",
"element": "red circular badge with text VS",
"connections": "blue arrow pointing left, orange arrow pointing right"
},
{
"position": "footer left",
"background": "light blue",
"text": "Dense: すべての層が常時稼働し全パラメータを使用",
"icon": "1 CPU chip graphic",
"label": "高い消費電力"
},
{
"position": "footer right",
"background": "light orange",
"text": "MoE: 必要なエキスパートのみを動員",
"icons": "2 circular graphics (orange arrows, blue lightning bolt)",
"label": "低コスト・高効率"
}
]
}
}