liqiang
b119135836
Build latest book artifacts / build (push) Canceled after 0s
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s
2225 lines
53 KiB
HTML
2225 lines
53 KiB
HTML
<!DOCTYPE html>
|
|
<html lang="en">
|
|
<head>
|
|
<meta charset="UTF-8">
|
|
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
|
<title>Model Leaderboard Evolution</title>
|
|
<script src="https://d3js.org/d3.v7.min.js"></script>
|
|
<style>
|
|
body {
|
|
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
|
margin: 0;
|
|
padding: 20px;
|
|
background: #f5f5f5;
|
|
}
|
|
|
|
#container {
|
|
max-width: 1200px;
|
|
margin: 0 auto;
|
|
background: white;
|
|
padding: 30px;
|
|
border-radius: 10px;
|
|
box-shadow: 0 2px 10px rgba(0,0,0,0.1);
|
|
}
|
|
|
|
h1 {
|
|
text-align: center;
|
|
color: #333;
|
|
margin-bottom: 10px;
|
|
}
|
|
|
|
#date-display {
|
|
text-align: center;
|
|
font-size: 24px;
|
|
font-weight: bold;
|
|
color: #666;
|
|
margin-bottom: 20px;
|
|
}
|
|
|
|
#chart {
|
|
margin: 20px 0;
|
|
}
|
|
|
|
.bar {
|
|
fill: steelblue;
|
|
cursor: pointer;
|
|
transition: fill 0.3s;
|
|
}
|
|
|
|
.bar:hover {
|
|
fill: #4682b4;
|
|
}
|
|
|
|
.bar-label {
|
|
font-size: 14px;
|
|
fill: white;
|
|
font-weight: bold;
|
|
}
|
|
|
|
.bar-value {
|
|
font-size: 12px;
|
|
fill: #333;
|
|
}
|
|
|
|
.rank-label {
|
|
font-size: 18px;
|
|
fill: #666;
|
|
font-weight: bold;
|
|
}
|
|
|
|
#controls {
|
|
text-align: center;
|
|
margin-top: 30px;
|
|
}
|
|
|
|
button {
|
|
padding: 10px 20px;
|
|
margin: 0 5px;
|
|
font-size: 16px;
|
|
cursor: pointer;
|
|
border: none;
|
|
border-radius: 5px;
|
|
background: #4CAF50;
|
|
color: white;
|
|
transition: background 0.3s;
|
|
}
|
|
|
|
button:hover {
|
|
background: #45a049;
|
|
}
|
|
|
|
button:disabled {
|
|
background: #ccc;
|
|
cursor: not-allowed;
|
|
}
|
|
|
|
#progress-bar {
|
|
width: 100%;
|
|
height: 5px;
|
|
background: #e0e0e0;
|
|
margin-top: 20px;
|
|
border-radius: 3px;
|
|
overflow: hidden;
|
|
}
|
|
|
|
#progress {
|
|
height: 100%;
|
|
background: #4CAF50;
|
|
width: 0%;
|
|
transition: width 0.5s;
|
|
}
|
|
|
|
#speed-control {
|
|
margin-top: 20px;
|
|
text-align: center;
|
|
}
|
|
|
|
#speed-slider {
|
|
width: 300px;
|
|
margin: 0 10px;
|
|
}
|
|
|
|
.info-box {
|
|
background: #f9f9f9;
|
|
padding: 15px;
|
|
border-radius: 5px;
|
|
margin-top: 20px;
|
|
font-size: 14px;
|
|
color: #666;
|
|
}
|
|
</style>
|
|
</head>
|
|
<body>
|
|
<div id="container">
|
|
<h1>🏆 Model Leaderboard Evolution</h1>
|
|
<div id="date-display">Loading...</div>
|
|
<div id="chart"></div>
|
|
<div id="controls">
|
|
<button id="play-btn">▶ Play</button>
|
|
<button id="pause-btn" disabled>⏸ Pause</button>
|
|
<button id="reset-btn">↺ Reset</button>
|
|
</div>
|
|
<div id="progress-bar">
|
|
<div id="progress"></div>
|
|
</div>
|
|
<div id="speed-control">
|
|
<label>Speed: </label>
|
|
<input type="range" id="speed-slider" min="1" max="10" value="5">
|
|
<span id="speed-value">5x</span>
|
|
</div>
|
|
<div class="info-box">
|
|
<strong>About:</strong> This animation shows the evolution of model rankings based on Elo ratings
|
|
calculated from Chatbot Arena voting data. Each frame represents a snapshot in time,
|
|
with models ranked by their current Elo rating. Bars show the rating value,
|
|
and the animation reveals how models compete and evolve over time.
|
|
</div>
|
|
</div>
|
|
|
|
<script>
|
|
const data = {
|
|
"frames": [
|
|
{
|
|
"date": "2023-04-30",
|
|
"timestamp": 1682898502,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "vicuna-13b",
|
|
"rating": 1148.0320274844,
|
|
"matches": 1793,
|
|
"wins": 1263.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "koala-13b",
|
|
"rating": 1094.6566198066,
|
|
"matches": 1304,
|
|
"wins": 712.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "oasst-pythia-12b",
|
|
"rating": 1020.2177219668,
|
|
"matches": 962,
|
|
"wins": 435.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "alpaca-13b",
|
|
"rating": 1010.2120122419,
|
|
"matches": 789,
|
|
"wins": 391.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "RWKV-4-Raven-14B",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "olmo-7b-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "openchat-3.5",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "openchat-3.5-0106",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "openhermes-2.5-mistral-7b",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "palm-2",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "phi-3-medium-4k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "phi-3-mini-128k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "phi-3-mini-4k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "phi-3-mini-4k-instruct-june-2024",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "phi-3-small-8k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-05-31",
|
|
"timestamp": 1685577242,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1238.9961322413,
|
|
"matches": 4262,
|
|
"wins": 3377.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "claude-1",
|
|
"rating": 1187.4877395882,
|
|
"matches": 3904,
|
|
"wins": 2830.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "claude-instant-1",
|
|
"rating": 1154.274136195,
|
|
"matches": 1952,
|
|
"wins": 1349.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1105.8720528902,
|
|
"matches": 4690,
|
|
"wins": 3188.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "vicuna-13b",
|
|
"rating": 1022.6112034193,
|
|
"matches": 6303,
|
|
"wins": 3850.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "palm-2",
|
|
"rating": 1016.3107510269,
|
|
"matches": 2557,
|
|
"wins": 1329.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "koala-13b",
|
|
"rating": 1012.4454280512,
|
|
"matches": 5844,
|
|
"wins": 2867.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "vicuna-7b",
|
|
"rating": 1011.0324202576,
|
|
"matches": 2150,
|
|
"wins": 998.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "olmo-7b-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "openchat-3.5",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "openchat-3.5-0106",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "llama-3-70b-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "phi-3-medium-4k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "phi-3-mini-128k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "phi-3-mini-4k-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-06-30",
|
|
"timestamp": 1688168016,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1202.7855178994,
|
|
"matches": 5138,
|
|
"wins": 4028.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "claude-1",
|
|
"rating": 1180.3356169594,
|
|
"matches": 4908,
|
|
"wins": 3549.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "claude-instant-1",
|
|
"rating": 1156.416237613,
|
|
"matches": 3348,
|
|
"wins": 2313.5
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1092.8321842502,
|
|
"matches": 130,
|
|
"wins": 95.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "vicuna-33b",
|
|
"rating": 1089.5376908161,
|
|
"matches": 585,
|
|
"wins": 355.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "vicuna-13b",
|
|
"rating": 1066.2818140283,
|
|
"matches": 7319,
|
|
"wins": 4391.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "mpt-30b-chat",
|
|
"rating": 1044.2195743582,
|
|
"matches": 581,
|
|
"wins": 309.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1038.8389236607,
|
|
"matches": 120,
|
|
"wins": 70.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "vicuna-7b",
|
|
"rating": 1036.034998319,
|
|
"matches": 3593,
|
|
"wins": 1688.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "wizardlm-13b",
|
|
"rating": 1035.7704972666,
|
|
"matches": 1751,
|
|
"wins": 944.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "guanaco-33b",
|
|
"rating": 1002.1886842778,
|
|
"matches": 1601,
|
|
"wins": 853.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "openhermes-2.5-mistral-7b",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "openchat-3.5-0106",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "openchat-3.5",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-07-31",
|
|
"timestamp": 1690847905,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "claude-1",
|
|
"rating": 1210.2521771739,
|
|
"matches": 5243,
|
|
"wins": 3793.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1202.7855178994,
|
|
"matches": 5138,
|
|
"wins": 4028.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1154.9674344467,
|
|
"matches": 950,
|
|
"wins": 672.5
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1140.353404289,
|
|
"matches": 907,
|
|
"wins": 595.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "claude-2.0",
|
|
"rating": 1118.7843220957,
|
|
"matches": 1244,
|
|
"wins": 779.5
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "claude-instant-1",
|
|
"rating": 1105.3005598323,
|
|
"matches": 4081,
|
|
"wins": 2766.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "vicuna-33b",
|
|
"rating": 1098.5182000539,
|
|
"matches": 1838,
|
|
"wins": 1123.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "mpt-30b-chat",
|
|
"rating": 1034.8806463899,
|
|
"matches": 1973,
|
|
"wins": 1040.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "wizardlm-13b",
|
|
"rating": 1033.5179170548,
|
|
"matches": 2892,
|
|
"wins": 1535.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "vicuna-13b",
|
|
"rating": 1023.9076403271,
|
|
"matches": 8087,
|
|
"wins": 4767.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "vicuna-7b",
|
|
"rating": 1016.1770238052,
|
|
"matches": 4161,
|
|
"wins": 1961.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "guanaco-33b",
|
|
"rating": 1002.9365315328,
|
|
"matches": 2796,
|
|
"wins": 1447.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "olmo-7b-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "llama-3-70b-instruct",
|
|
"rating": 1000.0,
|
|
"matches": 0,
|
|
"wins": 0.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-08-31",
|
|
"timestamp": 1693526088,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "claude-1",
|
|
"rating": 1210.2521771739,
|
|
"matches": 5243,
|
|
"wins": 3793.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1202.7855178994,
|
|
"matches": 5138,
|
|
"wins": 4028.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1119.8871952476,
|
|
"matches": 1665,
|
|
"wins": 1126.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1106.9846926653,
|
|
"matches": 1657,
|
|
"wins": 1028.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "claude-instant-1",
|
|
"rating": 1099.9530134295,
|
|
"matches": 4811,
|
|
"wins": 3188.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "vicuna-33b",
|
|
"rating": 1098.1637867641,
|
|
"matches": 2396,
|
|
"wins": 1440.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-2.0",
|
|
"rating": 1086.4442799903,
|
|
"matches": 2306,
|
|
"wins": 1424.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "wizardlm-13b",
|
|
"rating": 1083.2728493188,
|
|
"matches": 3430,
|
|
"wins": 1818.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "vicuna-13b",
|
|
"rating": 1033.4735754499,
|
|
"matches": 8726,
|
|
"wins": 5063.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "guanaco-33b",
|
|
"rating": 1027.2582838587,
|
|
"matches": 3003,
|
|
"wins": 1552.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "llama-2-13b-chat",
|
|
"rating": 1023.7199267566,
|
|
"matches": 2092,
|
|
"wins": 888.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "llama-2-7b-chat",
|
|
"rating": 1020.9720111389,
|
|
"matches": 1461,
|
|
"wins": 588.5
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "codellama-34b-instruct",
|
|
"rating": 1017.9060576554,
|
|
"matches": 297,
|
|
"wins": 133.5
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "mpt-30b-chat",
|
|
"rating": 1012.0958340026,
|
|
"matches": 2513,
|
|
"wins": 1281.5
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-09-30",
|
|
"timestamp": 1696118379,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "claude-1",
|
|
"rating": 1210.2521771739,
|
|
"matches": 5243,
|
|
"wins": 3793.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1202.7855178994,
|
|
"matches": 5138,
|
|
"wins": 4028.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1146.6804590093,
|
|
"matches": 3291,
|
|
"wins": 2142.5
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "claude-2.0",
|
|
"rating": 1131.051530135,
|
|
"matches": 3919,
|
|
"wins": 2397.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "wizardlm-70b",
|
|
"rating": 1099.8255313566,
|
|
"matches": 2860,
|
|
"wins": 1628.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "vicuna-33b",
|
|
"rating": 1089.9369903449,
|
|
"matches": 3519,
|
|
"wins": 2047.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1089.0381394624,
|
|
"matches": 3342,
|
|
"wins": 1979.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "claude-instant-1",
|
|
"rating": 1064.1008548893,
|
|
"matches": 6458,
|
|
"wins": 4068.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "wizardlm-13b",
|
|
"rating": 1041.6783008045,
|
|
"matches": 4637,
|
|
"wins": 2409.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "llama-2-13b-chat",
|
|
"rating": 1037.8614571446,
|
|
"matches": 3440,
|
|
"wins": 1520.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "guanaco-33b",
|
|
"rating": 1027.2582838587,
|
|
"matches": 3003,
|
|
"wins": 1552.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "mpt-30b-chat",
|
|
"rating": 1024.0563241095,
|
|
"matches": 2650,
|
|
"wins": 1351.5
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "llama-2-70b-chat",
|
|
"rating": 1018.3776994547,
|
|
"matches": 3075,
|
|
"wins": 1517.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "codellama-34b-instruct",
|
|
"rating": 1015.6238126357,
|
|
"matches": 2326,
|
|
"wins": 1011.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-10-31",
|
|
"timestamp": 1698796794,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1202.7855178994,
|
|
"matches": 5138,
|
|
"wins": 4028.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "claude-1",
|
|
"rating": 1135.2779227948,
|
|
"matches": 6216,
|
|
"wins": 4349.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1123.7060081331,
|
|
"matches": 5209,
|
|
"wins": 3288.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "claude-2.0",
|
|
"rating": 1119.1576056023,
|
|
"matches": 6036,
|
|
"wins": 3590.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "llama-2-70b-chat",
|
|
"rating": 1087.7175914837,
|
|
"matches": 4471,
|
|
"wins": 2264.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1085.1368261005,
|
|
"matches": 5976,
|
|
"wins": 3407.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "llama-2-7b-chat",
|
|
"rating": 1062.7898867404,
|
|
"matches": 3390,
|
|
"wins": 1445.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "llama-2-13b-chat",
|
|
"rating": 1055.1004769603,
|
|
"matches": 5256,
|
|
"wins": 2402.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "claude-instant-1",
|
|
"rating": 1052.5857472734,
|
|
"matches": 7963,
|
|
"wins": 4853.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "vicuna-33b",
|
|
"rating": 1048.8668082886,
|
|
"matches": 4691,
|
|
"wins": 2669.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "wizardlm-13b",
|
|
"rating": 1045.8352687204,
|
|
"matches": 5834,
|
|
"wins": 3039.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "zephyr-7b-alpha",
|
|
"rating": 1037.538873926,
|
|
"matches": 1406,
|
|
"wins": 670.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "zephyr-7b-beta",
|
|
"rating": 1032.3072561544,
|
|
"matches": 2427,
|
|
"wins": 1188.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "guanaco-33b",
|
|
"rating": 1027.2582838587,
|
|
"matches": 3003,
|
|
"wins": 1552.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-11-30",
|
|
"timestamp": 1701388447,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1202.7855178994,
|
|
"matches": 5138,
|
|
"wins": 4028.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1198.6520935367,
|
|
"matches": 5841,
|
|
"wins": 3848.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1136.1538904497,
|
|
"matches": 10479,
|
|
"wins": 6181.5
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "claude-2.1",
|
|
"rating": 1110.951707598,
|
|
"matches": 3793,
|
|
"wins": 1867.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "claude-2.0",
|
|
"rating": 1099.6998763487,
|
|
"matches": 8802,
|
|
"wins": 4961.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "claude-1",
|
|
"rating": 1098.7945603749,
|
|
"matches": 8703,
|
|
"wins": 5713.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-instant-1",
|
|
"rating": 1088.8448799088,
|
|
"matches": 10414,
|
|
"wins": 6125.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "tulu-2-dpo-70b",
|
|
"rating": 1070.1394385,
|
|
"matches": 1838,
|
|
"wins": 938.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "wizardlm-70b",
|
|
"rating": 1067.5699214813,
|
|
"matches": 4791,
|
|
"wins": 2556.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "openhermes-2.5-mistral-7b",
|
|
"rating": 1041.9395267947,
|
|
"matches": 462,
|
|
"wins": 225.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1038.6242581712,
|
|
"matches": 13457,
|
|
"wins": 7256.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "openchat-3.5",
|
|
"rating": 1037.921915116,
|
|
"matches": 4071,
|
|
"wins": 1969.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "yi-34b-chat",
|
|
"rating": 1035.0054237172,
|
|
"matches": 1999,
|
|
"wins": 1018.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "guanaco-33b",
|
|
"rating": 1027.2582838587,
|
|
"matches": 3003,
|
|
"wins": 1552.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2023-12-31",
|
|
"timestamp": 1704067179,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1217.102910309,
|
|
"matches": 17838,
|
|
"wins": 12277.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1122.8848157788,
|
|
"matches": 14270,
|
|
"wins": 9300.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "mistral-medium",
|
|
"rating": 1113.6768319273,
|
|
"matches": 2022,
|
|
"wins": 1066.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1103.0918923988,
|
|
"matches": 16160,
|
|
"wins": 9157.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "gemini-pro",
|
|
"rating": 1087.2299210223,
|
|
"matches": 6335,
|
|
"wins": 2928.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "claude-2.1",
|
|
"rating": 1075.8091604699,
|
|
"matches": 17783,
|
|
"wins": 8358.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-1",
|
|
"rating": 1074.6731858986,
|
|
"matches": 15697,
|
|
"wins": 9408.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "claude-2.0",
|
|
"rating": 1074.3390927875,
|
|
"matches": 10144,
|
|
"wins": 5596.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "mixtral-8x7b-instruct-v0.1",
|
|
"rating": 1070.0043673617,
|
|
"matches": 10665,
|
|
"wins": 5428.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "wizardlm-70b",
|
|
"rating": 1067.6969362061,
|
|
"matches": 6668,
|
|
"wins": 3504.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1058.3556624364,
|
|
"matches": 22731,
|
|
"wins": 11800.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "claude-instant-1",
|
|
"rating": 1057.7265161573,
|
|
"matches": 14379,
|
|
"wins": 8096.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "llama-2-70b-chat",
|
|
"rating": 1054.9183137869,
|
|
"matches": 11874,
|
|
"wins": 5952.5
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "vicuna-33b",
|
|
"rating": 1049.2500592227,
|
|
"matches": 13805,
|
|
"wins": 7138.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-01-31",
|
|
"timestamp": 1706745570,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1191.7291570449,
|
|
"matches": 2995,
|
|
"wins": 1934.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1177.1753011023,
|
|
"matches": 32110,
|
|
"wins": 21947.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1156.0223779532,
|
|
"matches": 8180,
|
|
"wins": 5087.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1102.6110894791,
|
|
"matches": 17618,
|
|
"wins": 11130.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1096.5543629849,
|
|
"matches": 27503,
|
|
"wins": 14889.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "mistral-medium",
|
|
"rating": 1088.1364842608,
|
|
"matches": 13766,
|
|
"wins": 7277.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-2.0",
|
|
"rating": 1078.6576791259,
|
|
"matches": 11196,
|
|
"wins": 6079.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "claude-1",
|
|
"rating": 1074.6005861725,
|
|
"matches": 17178,
|
|
"wins": 10192.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "claude-instant-1",
|
|
"rating": 1074.2260629955,
|
|
"matches": 17272,
|
|
"wins": 9516.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "yi-34b-chat",
|
|
"rating": 1070.0428903174,
|
|
"matches": 6182,
|
|
"wins": 3245.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "gemini-pro",
|
|
"rating": 1059.2020149016,
|
|
"matches": 6565,
|
|
"wins": 3028.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "gemini-pro-dev-api",
|
|
"rating": 1056.048106887,
|
|
"matches": 6947,
|
|
"wins": 3136.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "starling-lm-7b-alpha",
|
|
"rating": 1053.4076520126,
|
|
"matches": 4754,
|
|
"wins": 2306.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "llama-2-13b-chat",
|
|
"rating": 1044.1644167527,
|
|
"matches": 11049,
|
|
"wins": 5241.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-02-29",
|
|
"timestamp": 1709251194,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1208.9204699264,
|
|
"matches": 40506,
|
|
"wins": 27643.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1180.9617003686,
|
|
"matches": 9465,
|
|
"wins": 6303.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1124.025205285,
|
|
"matches": 34716,
|
|
"wins": 18806.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "mistral-large-2402",
|
|
"rating": 1108.4940445417,
|
|
"matches": 2204,
|
|
"wins": 1230.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "claude-1",
|
|
"rating": 1093.8638255361,
|
|
"matches": 19255,
|
|
"wins": 11318.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1091.9146380654,
|
|
"matches": 21597,
|
|
"wins": 13396.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "mistral-medium",
|
|
"rating": 1090.7368867398,
|
|
"matches": 19606,
|
|
"wins": 10338.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "qwen1.5-72b-chat",
|
|
"rating": 1084.6946494744,
|
|
"matches": 12900,
|
|
"wins": 6947.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1082.4079728766,
|
|
"matches": 11394,
|
|
"wins": 6848.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "gemini-pro-dev-api",
|
|
"rating": 1080.9212410838,
|
|
"matches": 10220,
|
|
"wins": 4711.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "mistral-next",
|
|
"rating": 1066.1598337751,
|
|
"matches": 11142,
|
|
"wins": 5435.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "claude-2.0",
|
|
"rating": 1065.9386839783,
|
|
"matches": 12730,
|
|
"wins": 6794.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "gpt-3.5-turbo-0613",
|
|
"rating": 1063.962757939,
|
|
"matches": 37529,
|
|
"wins": 18696.5
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "gemini-pro",
|
|
"rating": 1059.2020149016,
|
|
"matches": 6565,
|
|
"wins": 3028.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-03-31",
|
|
"timestamp": 1711929587,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "claude-3-opus-20240229",
|
|
"rating": 1176.1614412178,
|
|
"matches": 37154,
|
|
"wins": 23362.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1167.8343686963,
|
|
"matches": 37927,
|
|
"wins": 23595.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1154.534750616,
|
|
"matches": 55093,
|
|
"wins": 36153.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1118.0101239811,
|
|
"matches": 34923,
|
|
"wins": 20378.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "gpt-4-0613",
|
|
"rating": 1102.2112432541,
|
|
"matches": 52660,
|
|
"wins": 26777.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "claude-3-haiku-20240307",
|
|
"rating": 1098.4007222704,
|
|
"matches": 29159,
|
|
"wins": 15336.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1094.2118678517,
|
|
"matches": 11827,
|
|
"wins": 7075.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "claude-1",
|
|
"rating": 1092.7309868534,
|
|
"matches": 21172,
|
|
"wins": 12081.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "claude-3-sonnet-20240229",
|
|
"rating": 1091.7321545888,
|
|
"matches": 42016,
|
|
"wins": 23536.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "mistral-large-2402",
|
|
"rating": 1086.4200447987,
|
|
"matches": 28004,
|
|
"wins": 13921.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "qwen1.5-72b-chat",
|
|
"rating": 1080.8057648949,
|
|
"matches": 21603,
|
|
"wins": 11268.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "gemini-pro-dev-api",
|
|
"rating": 1071.1907344599,
|
|
"matches": 15720,
|
|
"wins": 7294.5
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "claude-2.0",
|
|
"rating": 1069.530762176,
|
|
"matches": 12774,
|
|
"wins": 6814.5
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "gemini-pro",
|
|
"rating": 1059.2020149016,
|
|
"matches": 6565,
|
|
"wins": 3028.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-04-30",
|
|
"timestamp": 1714521598,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "claude-3-opus-20240229",
|
|
"rating": 1174.9545916992,
|
|
"matches": 73578,
|
|
"wins": 45143.0
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1161.5891307634,
|
|
"matches": 60447,
|
|
"wins": 36723.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4-turbo-2024-04-09",
|
|
"rating": 1156.3586862763,
|
|
"matches": 30972,
|
|
"wins": 18505.5
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gemini-1.5-pro-api-0409-preview",
|
|
"rating": 1154.3497993228,
|
|
"matches": 34072,
|
|
"wins": 20486.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1145.9992036875,
|
|
"matches": 67855,
|
|
"wins": 43747.5
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "llama-3-70b-instruct",
|
|
"rating": 1128.5863503665,
|
|
"matches": 41376,
|
|
"wins": 23088.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-3-sonnet-20240229",
|
|
"rating": 1103.1308363864,
|
|
"matches": 71239,
|
|
"wins": 39286.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "gpt-4-0314",
|
|
"rating": 1099.6764239908,
|
|
"matches": 47325,
|
|
"wins": 26914.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1094.2118678517,
|
|
"matches": 11827,
|
|
"wins": 7075.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "claude-1",
|
|
"rating": 1092.7309868534,
|
|
"matches": 21172,
|
|
"wins": 12081.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "gemini-pro-dev-api",
|
|
"rating": 1072.3072263884,
|
|
"matches": 18790,
|
|
"wins": 8728.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "claude-2.0",
|
|
"rating": 1069.530762176,
|
|
"matches": 12774,
|
|
"wins": 6814.5
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "mistral-large-2402",
|
|
"rating": 1063.9233829534,
|
|
"matches": 42913,
|
|
"wins": 21064.5
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "mistral-medium",
|
|
"rating": 1060.0648399023,
|
|
"matches": 31488,
|
|
"wins": 15950.0
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-05-31",
|
|
"timestamp": 1717199998,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gemini-1.5-pro-api-0409-preview",
|
|
"rating": 1159.8203895968,
|
|
"matches": 55654,
|
|
"wins": 33738.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gemini-1.5-pro-api-0514",
|
|
"rating": 1143.8706808856,
|
|
"matches": 24524,
|
|
"wins": 14254.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4o-2024-05-13",
|
|
"rating": 1137.3671645518,
|
|
"matches": 31160,
|
|
"wins": 19149.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-4-turbo-2024-04-09",
|
|
"rating": 1118.9995587843,
|
|
"matches": 59629,
|
|
"wins": 34823.5
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gemini-advanced-0514",
|
|
"rating": 1117.3066177883,
|
|
"matches": 26204,
|
|
"wins": 15262.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1112.5274461375,
|
|
"matches": 79593,
|
|
"wins": 50441.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-3-opus-20240229",
|
|
"rating": 1101.7492312993,
|
|
"matches": 122485,
|
|
"wins": 72927.0
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1094.2118678517,
|
|
"matches": 11827,
|
|
"wins": 7075.5
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "claude-1",
|
|
"rating": 1092.7309868534,
|
|
"matches": 21172,
|
|
"wins": 12081.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1081.2915676288,
|
|
"matches": 72771,
|
|
"wins": 43712.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "gemini-1.5-flash-api-0514",
|
|
"rating": 1080.7471858528,
|
|
"matches": 22725,
|
|
"wins": 12322.0
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "llama-3-70b-instruct",
|
|
"rating": 1080.1146241575,
|
|
"matches": 123617,
|
|
"wins": 66125.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "yi-large-preview",
|
|
"rating": 1075.1581477738,
|
|
"matches": 33808,
|
|
"wins": 18820.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "claude-3-sonnet-20240229",
|
|
"rating": 1073.0442365702,
|
|
"matches": 95310,
|
|
"wins": 51518.5
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-06-30",
|
|
"timestamp": 1719791933,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gemini-1.5-pro-api-0409-preview",
|
|
"rating": 1159.8203895968,
|
|
"matches": 55654,
|
|
"wins": 33738.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gpt-4o-2024-05-13",
|
|
"rating": 1133.8475199011,
|
|
"matches": 51867,
|
|
"wins": 32040.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "claude-3-5-sonnet-20240620",
|
|
"rating": 1128.0631691547,
|
|
"matches": 19006,
|
|
"wins": 11116.5
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gemini-advanced-0514",
|
|
"rating": 1104.3737643515,
|
|
"matches": 40045,
|
|
"wins": 23603.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "gemini-1.5-pro-api-0514",
|
|
"rating": 1102.3509445504,
|
|
"matches": 44696,
|
|
"wins": 25947.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1094.2118678517,
|
|
"matches": 11827,
|
|
"wins": 7075.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-1",
|
|
"rating": 1092.7309868534,
|
|
"matches": 21172,
|
|
"wins": 12081.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "claude-3-opus-20240229",
|
|
"rating": 1089.7718969029,
|
|
"matches": 140092,
|
|
"wins": 83000.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "gpt-4-turbo-2024-04-09",
|
|
"rating": 1087.4030763635,
|
|
"matches": 70553,
|
|
"wins": 41271.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1085.683304997,
|
|
"matches": 85436,
|
|
"wins": 53702.5
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "yi-large",
|
|
"rating": 1084.1099178358,
|
|
"matches": 10333,
|
|
"wins": 5414.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "yi-large-preview",
|
|
"rating": 1082.5341263969,
|
|
"matches": 45552,
|
|
"wins": 25462.5
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1077.8233462851,
|
|
"matches": 78726,
|
|
"wins": 47042.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "gemini-pro-dev-api",
|
|
"rating": 1072.3072263884,
|
|
"matches": 18790,
|
|
"wins": 8728.5
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-07-31",
|
|
"timestamp": 1722470391,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gemini-1.5-pro-api-0409-preview",
|
|
"rating": 1159.8203895968,
|
|
"matches": 55654,
|
|
"wins": 33738.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gemini-1.5-pro-exp-0801",
|
|
"rating": 1126.0573644679,
|
|
"matches": 12851,
|
|
"wins": 7947.0
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "gpt-4o-mini-2024-07-18",
|
|
"rating": 1114.5858560551,
|
|
"matches": 12601,
|
|
"wins": 7389.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "gpt-4o-2024-05-13",
|
|
"rating": 1101.9127952885,
|
|
"matches": 70146,
|
|
"wins": 42771.5
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "claude-3-5-sonnet-20240620",
|
|
"rating": 1095.983029238,
|
|
"matches": 40495,
|
|
"wins": 23176.0
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1094.2118678517,
|
|
"matches": 11827,
|
|
"wins": 7075.5
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "claude-1",
|
|
"rating": 1092.7309868534,
|
|
"matches": 21172,
|
|
"wins": 12081.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "gpt-4-0125-preview",
|
|
"rating": 1090.9725236983,
|
|
"matches": 83410,
|
|
"wins": 49571.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "llama-3.1-405b-instruct",
|
|
"rating": 1089.0830838178,
|
|
"matches": 9068,
|
|
"wins": 5012.5
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "gpt-4-1106-preview",
|
|
"rating": 1088.0773741892,
|
|
"matches": 90079,
|
|
"wins": 56256.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "gemini-advanced-0514",
|
|
"rating": 1086.0595644473,
|
|
"matches": 50857,
|
|
"wins": 29822.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "athene-70b-0725",
|
|
"rating": 1083.33516485,
|
|
"matches": 5707,
|
|
"wins": 3074.0
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "mistral-large-2407",
|
|
"rating": 1083.0034192178,
|
|
"matches": 3410,
|
|
"wins": 1798.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "gemini-1.5-pro-api-0514",
|
|
"rating": 1079.0286607691,
|
|
"matches": 62293,
|
|
"wins": 35591.5
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"date": "2024-08-14",
|
|
"timestamp": 1723665173,
|
|
"models": [
|
|
{
|
|
"rank": 1,
|
|
"name": "gemini-1.5-pro-api-0409-preview",
|
|
"rating": 1159.8203895968,
|
|
"matches": 55654,
|
|
"wins": 33738.5
|
|
},
|
|
{
|
|
"rank": 2,
|
|
"name": "gemini-1.5-pro-exp-0801",
|
|
"rating": 1127.8877886511,
|
|
"matches": 20071,
|
|
"wins": 12139.5
|
|
},
|
|
{
|
|
"rank": 3,
|
|
"name": "chatgpt-4o-latest",
|
|
"rating": 1126.6935165964,
|
|
"matches": 14514,
|
|
"wins": 8999.0
|
|
},
|
|
{
|
|
"rank": 4,
|
|
"name": "gpt-3.5-turbo-0314",
|
|
"rating": 1103.9751202512,
|
|
"matches": 5655,
|
|
"wins": 3789.0
|
|
},
|
|
{
|
|
"rank": 5,
|
|
"name": "bard-jan-24-gemini-pro",
|
|
"rating": 1094.2118678517,
|
|
"matches": 11827,
|
|
"wins": 7075.5
|
|
},
|
|
{
|
|
"rank": 6,
|
|
"name": "claude-1",
|
|
"rating": 1092.7309868534,
|
|
"matches": 21172,
|
|
"wins": 12081.5
|
|
},
|
|
{
|
|
"rank": 7,
|
|
"name": "gemini-advanced-0514",
|
|
"rating": 1087.5408038371,
|
|
"matches": 52155,
|
|
"wins": 30523.0
|
|
},
|
|
{
|
|
"rank": 8,
|
|
"name": "llama-3.1-70b-instruct",
|
|
"rating": 1087.3529845908,
|
|
"matches": 13604,
|
|
"wins": 7058.5
|
|
},
|
|
{
|
|
"rank": 9,
|
|
"name": "gpt-4o-2024-05-13",
|
|
"rating": 1086.7242770205,
|
|
"matches": 77509,
|
|
"wins": 46951.0
|
|
},
|
|
{
|
|
"rank": 10,
|
|
"name": "gpt-4o-2024-08-06",
|
|
"rating": 1085.601246973,
|
|
"matches": 9761,
|
|
"wins": 5323.0
|
|
},
|
|
{
|
|
"rank": 11,
|
|
"name": "gpt-4o-mini-2024-07-18",
|
|
"rating": 1079.187026245,
|
|
"matches": 19370,
|
|
"wins": 11082.0
|
|
},
|
|
{
|
|
"rank": 12,
|
|
"name": "claude-3-5-sonnet-20240620",
|
|
"rating": 1079.0096726066,
|
|
"matches": 47703,
|
|
"wins": 27107.5
|
|
},
|
|
{
|
|
"rank": 13,
|
|
"name": "claude-3-opus-20240229",
|
|
"rating": 1074.9078256001,
|
|
"matches": 155944,
|
|
"wins": 91432.5
|
|
},
|
|
{
|
|
"rank": 14,
|
|
"name": "athene-70b-0725",
|
|
"rating": 1073.7607622933,
|
|
"matches": 11476,
|
|
"wins": 6098.0
|
|
},
|
|
{
|
|
"rank": 15,
|
|
"name": "gemini-pro-dev-api",
|
|
"rating": 1072.3072263884,
|
|
"matches": 18790,
|
|
"wins": 8728.5
|
|
}
|
|
]
|
|
}
|
|
],
|
|
"total_frames": 17,
|
|
"top_n": 15,
|
|
"start_date": "2023-04-30",
|
|
"end_date": "2024-08-14"
|
|
};
|
|
|
|
// Configuration
|
|
const margin = {top: 20, right: 100, bottom: 40, left: 50};
|
|
const width = 1100 - margin.left - margin.right;
|
|
const height = 600 - margin.top - margin.bottom;
|
|
const barHeight = height / data.top_n - 5;
|
|
|
|
// Create SVG
|
|
const svg = d3.select("#chart")
|
|
.append("svg")
|
|
.attr("width", width + margin.left + margin.right)
|
|
.attr("height", height + margin.top + margin.bottom)
|
|
.append("g")
|
|
.attr("transform", `translate(${margin.left},${margin.top})`);
|
|
|
|
// Scales
|
|
const xScale = d3.scaleLinear()
|
|
.domain([0, d3.max(data.frames.flatMap(f => f.models.map(m => m.rating)))])
|
|
.range([0, width - 200]);
|
|
|
|
// Color scale
|
|
const colorScale = d3.scaleOrdinal(d3.schemeCategory10);
|
|
|
|
// Animation state
|
|
let currentFrame = 0;
|
|
let isPlaying = false;
|
|
let animationInterval = null;
|
|
let animationSpeed = 500; // milliseconds per frame
|
|
|
|
// Update speed based on slider
|
|
d3.select("#speed-slider").on("input", function() {
|
|
const speed = +this.value;
|
|
animationSpeed = 1000 / speed;
|
|
d3.select("#speed-value").text(`${speed}x`);
|
|
if (isPlaying) {
|
|
stopAnimation();
|
|
startAnimation();
|
|
}
|
|
});
|
|
|
|
function updateChart(frameIndex) {
|
|
const frame = data.frames[frameIndex];
|
|
|
|
// Update date display
|
|
d3.select("#date-display").text(frame.date);
|
|
|
|
// Update progress bar
|
|
const progress = ((frameIndex + 1) / data.total_frames) * 100;
|
|
d3.select("#progress").style("width", `${progress}%`);
|
|
|
|
// Update max value for scale
|
|
const maxRating = d3.max(frame.models, d => d.rating);
|
|
xScale.domain([0, maxRating * 1.1]);
|
|
|
|
// Bind data
|
|
const bars = svg.selectAll(".bar-group")
|
|
.data(frame.models, d => d.name);
|
|
|
|
// Remove old bars
|
|
bars.exit()
|
|
.transition()
|
|
.duration(animationSpeed * 0.8)
|
|
.style("opacity", 0)
|
|
.remove();
|
|
|
|
// Add new bars
|
|
const enter = bars.enter()
|
|
.append("g")
|
|
.attr("class", "bar-group")
|
|
.style("opacity", 0);
|
|
|
|
enter.append("rect")
|
|
.attr("class", "bar")
|
|
.attr("height", barHeight);
|
|
|
|
enter.append("text")
|
|
.attr("class", "bar-label")
|
|
.attr("x", 10)
|
|
.attr("y", barHeight / 2)
|
|
.attr("dy", "0.35em");
|
|
|
|
enter.append("text")
|
|
.attr("class", "bar-value")
|
|
.attr("y", barHeight / 2)
|
|
.attr("dy", "0.35em");
|
|
|
|
enter.append("text")
|
|
.attr("class", "rank-label")
|
|
.attr("x", -40)
|
|
.attr("y", barHeight / 2)
|
|
.attr("dy", "0.35em")
|
|
.attr("text-anchor", "middle");
|
|
|
|
// Update all bars
|
|
const merged = enter.merge(bars);
|
|
|
|
merged.transition()
|
|
.duration(animationSpeed * 0.8)
|
|
.style("opacity", 1)
|
|
.attr("transform", (d, i) => `translate(0,${i * (barHeight + 5)})`);
|
|
|
|
merged.select(".bar")
|
|
.transition()
|
|
.duration(animationSpeed * 0.8)
|
|
.attr("width", d => xScale(d.rating))
|
|
.attr("fill", d => colorScale(d.name));
|
|
|
|
merged.select(".bar-label")
|
|
.text(d => d.name);
|
|
|
|
merged.select(".bar-value")
|
|
.transition()
|
|
.duration(animationSpeed * 0.8)
|
|
.attr("x", d => xScale(d.rating) + 10)
|
|
.text(d => `${Math.round(d.rating)} (${d.matches} matches)`);
|
|
|
|
merged.select(".rank-label")
|
|
.text(d => `#${d.rank}`);
|
|
}
|
|
|
|
function startAnimation() {
|
|
if (currentFrame >= data.total_frames - 1) {
|
|
currentFrame = 0;
|
|
}
|
|
|
|
isPlaying = true;
|
|
d3.select("#play-btn").property("disabled", true);
|
|
d3.select("#pause-btn").property("disabled", false);
|
|
|
|
animationInterval = setInterval(() => {
|
|
updateChart(currentFrame);
|
|
currentFrame++;
|
|
|
|
if (currentFrame >= data.total_frames) {
|
|
stopAnimation();
|
|
currentFrame = data.total_frames - 1;
|
|
}
|
|
}, animationSpeed);
|
|
}
|
|
|
|
function stopAnimation() {
|
|
isPlaying = false;
|
|
d3.select("#play-btn").property("disabled", false);
|
|
d3.select("#pause-btn").property("disabled", true);
|
|
|
|
if (animationInterval) {
|
|
clearInterval(animationInterval);
|
|
animationInterval = null;
|
|
}
|
|
}
|
|
|
|
function resetAnimation() {
|
|
stopAnimation();
|
|
currentFrame = 0;
|
|
updateChart(currentFrame);
|
|
d3.select("#progress").style("width", "0%");
|
|
}
|
|
|
|
// Button handlers
|
|
d3.select("#play-btn").on("click", startAnimation);
|
|
d3.select("#pause-btn").on("click", stopAnimation);
|
|
d3.select("#reset-btn").on("click", resetAnimation);
|
|
|
|
// Initialize with first frame
|
|
updateChart(0);
|
|
|
|
// Auto-play on load
|
|
setTimeout(startAnimation, 1000);
|
|
</script>
|
|
</body>
|
|
</html>
|