|
| 1 | +import React from 'react'; |
| 2 | + |
| 3 | +/** |
| 4 | + * Static, theme-aware overview of the VLM-judge evaluation protocol: |
| 5 | + * a pair of meshes for the same input photo -> a fixed 24-view headless |
| 6 | + * render rig -> two *independent* VLM judge families, each queried in both |
| 7 | + * presentation orders -> keep only the order-consistent verdicts -> report |
| 8 | + * cross-model agreement. The cheap proxies are scored against the same |
| 9 | + * (de-biased) reference. |
| 10 | + * |
| 11 | + * Strokes and text use `currentColor` / Infima CSS vars so the figure adapts |
| 12 | + * to light and dark mode. The green accent marks the de-biased reference path. |
| 13 | + */ |
| 14 | +export default function JudgeProtocolDiagram(): JSX.Element { |
| 15 | + const ok = '#2e8b57'; |
| 16 | + |
| 17 | + // camera ring for the render rig, centred above the "24 views" label |
| 18 | + const ringCx = 270; |
| 19 | + const ringCy = 140; |
| 20 | + const ringR = 20; |
| 21 | + |
| 22 | + return ( |
| 23 | + <figure style={{ margin: '2rem 0' }}> |
| 24 | + <svg |
| 25 | + viewBox="0 0 820 320" |
| 26 | + role="img" |
| 27 | + aria-label="A mesh pair is rendered from 24 views, judged by two independent VLM families in both presentation orders, order-inconsistent verdicts are discarded, and cross-model agreement is reported; cheap proxies are scored against the same de-biased reference." |
| 28 | + style={{ |
| 29 | + width: '100%', |
| 30 | + height: 'auto', |
| 31 | + color: 'var(--ifm-font-color-base)', |
| 32 | + }} |
| 33 | + fill="none" |
| 34 | + > |
| 35 | + {/* Stage labels */} |
| 36 | + <g |
| 37 | + fontSize="12" |
| 38 | + fontWeight={600} |
| 39 | + textAnchor="middle" |
| 40 | + fill="currentColor" |
| 41 | + opacity={0.6} |
| 42 | + > |
| 43 | + <text x="80" y="24"> |
| 44 | + Mesh pair |
| 45 | + </text> |
| 46 | + <text x="270" y="24"> |
| 47 | + 24-view render rig |
| 48 | + </text> |
| 49 | + <text x="490" y="24"> |
| 50 | + Two independent judges |
| 51 | + </text> |
| 52 | + <text x="725" y="24"> |
| 53 | + De-biased verdict |
| 54 | + </text> |
| 55 | + </g> |
| 56 | + |
| 57 | + {/* Mesh pair box */} |
| 58 | + <rect |
| 59 | + x="20" |
| 60 | + y="110" |
| 61 | + width="120" |
| 62 | + height="80" |
| 63 | + rx="8" |
| 64 | + stroke="currentColor" |
| 65 | + strokeWidth={1.5} |
| 66 | + /> |
| 67 | + <text |
| 68 | + x="80" |
| 69 | + y="140" |
| 70 | + fontSize="13" |
| 71 | + textAnchor="middle" |
| 72 | + fill="currentColor" |
| 73 | + > |
| 74 | + mesh A |
| 75 | + </text> |
| 76 | + <text |
| 77 | + x="80" |
| 78 | + y="157" |
| 79 | + fontSize="11" |
| 80 | + textAnchor="middle" |
| 81 | + fill="currentColor" |
| 82 | + opacity={0.5} |
| 83 | + > |
| 84 | + vs |
| 85 | + </text> |
| 86 | + <text |
| 87 | + x="80" |
| 88 | + y="176" |
| 89 | + fontSize="13" |
| 90 | + textAnchor="middle" |
| 91 | + fill="currentColor" |
| 92 | + > |
| 93 | + mesh B |
| 94 | + </text> |
| 95 | + |
| 96 | + {/* arrow -> render rig */} |
| 97 | + <path |
| 98 | + d="M140 150 L196 150" |
| 99 | + stroke="currentColor" |
| 100 | + strokeWidth={1.25} |
| 101 | + opacity={0.5} |
| 102 | + markerEnd="url(#jp-arrow)" |
| 103 | + /> |
| 104 | + |
| 105 | + {/* Render rig box: camera ring on top, label below */} |
| 106 | + <rect |
| 107 | + x="200" |
| 108 | + y="110" |
| 109 | + width="140" |
| 110 | + height="80" |
| 111 | + rx="8" |
| 112 | + stroke="currentColor" |
| 113 | + strokeWidth={1.5} |
| 114 | + /> |
| 115 | + {[0, 45, 90, 135, 180, 225, 270, 315].map((deg) => { |
| 116 | + const rad = (deg * Math.PI) / 180; |
| 117 | + return ( |
| 118 | + <circle |
| 119 | + key={deg} |
| 120 | + cx={ringCx + ringR * Math.cos(rad)} |
| 121 | + cy={ringCy + ringR * Math.sin(rad)} |
| 122 | + r={2.5} |
| 123 | + fill="currentColor" |
| 124 | + opacity={0.55} |
| 125 | + /> |
| 126 | + ); |
| 127 | + })} |
| 128 | + <text |
| 129 | + x="270" |
| 130 | + y="182" |
| 131 | + fontSize="11" |
| 132 | + textAnchor="middle" |
| 133 | + fill="currentColor" |
| 134 | + opacity={0.6} |
| 135 | + > |
| 136 | + 24 views |
| 137 | + </text> |
| 138 | + |
| 139 | + {/* arrow -> judges */} |
| 140 | + <path |
| 141 | + d="M340 150 L396 150" |
| 142 | + stroke="currentColor" |
| 143 | + strokeWidth={1.25} |
| 144 | + opacity={0.5} |
| 145 | + markerEnd="url(#jp-arrow)" |
| 146 | + /> |
| 147 | + |
| 148 | + {/* Judge X */} |
| 149 | + <rect |
| 150 | + x="400" |
| 151 | + y="92" |
| 152 | + width="190" |
| 153 | + height="48" |
| 154 | + rx="8" |
| 155 | + stroke="currentColor" |
| 156 | + strokeWidth={1.5} |
| 157 | + /> |
| 158 | + <text |
| 159 | + x="495" |
| 160 | + y="112" |
| 161 | + fontSize="12" |
| 162 | + fontWeight={600} |
| 163 | + textAnchor="middle" |
| 164 | + fill="currentColor" |
| 165 | + > |
| 166 | + Judge X · Qwen2.5-VL-7B |
| 167 | + </text> |
| 168 | + <text |
| 169 | + x="495" |
| 170 | + y="129" |
| 171 | + fontSize="11" |
| 172 | + textAnchor="middle" |
| 173 | + fill="currentColor" |
| 174 | + opacity={0.6} |
| 175 | + > |
| 176 | + both orders: (A,B) and (B,A) |
| 177 | + </text> |
| 178 | + |
| 179 | + {/* Judge Y */} |
| 180 | + <rect |
| 181 | + x="400" |
| 182 | + y="160" |
| 183 | + width="190" |
| 184 | + height="48" |
| 185 | + rx="8" |
| 186 | + stroke="currentColor" |
| 187 | + strokeWidth={1.5} |
| 188 | + /> |
| 189 | + <text |
| 190 | + x="495" |
| 191 | + y="180" |
| 192 | + fontSize="12" |
| 193 | + fontWeight={600} |
| 194 | + textAnchor="middle" |
| 195 | + fill="currentColor" |
| 196 | + > |
| 197 | + Judge Y · InternVL3-8B |
| 198 | + </text> |
| 199 | + <text |
| 200 | + x="495" |
| 201 | + y="197" |
| 202 | + fontSize="11" |
| 203 | + textAnchor="middle" |
| 204 | + fill="currentColor" |
| 205 | + opacity={0.6} |
| 206 | + > |
| 207 | + independent validation |
| 208 | + </text> |
| 209 | + |
| 210 | + {/* arrows judges -> filter */} |
| 211 | + <path |
| 212 | + d="M590 116 C 626 116, 626 150, 656 150" |
| 213 | + stroke="currentColor" |
| 214 | + strokeWidth={1.25} |
| 215 | + opacity={0.5} |
| 216 | + markerEnd="url(#jp-arrow)" |
| 217 | + /> |
| 218 | + <path |
| 219 | + d="M590 184 C 626 184, 626 150, 656 150" |
| 220 | + stroke="currentColor" |
| 221 | + strokeWidth={1.25} |
| 222 | + opacity={0.5} |
| 223 | + markerEnd="url(#jp-arrow)" |
| 224 | + /> |
| 225 | + |
| 226 | + {/* Keep-consistent / agreement box */} |
| 227 | + <rect |
| 228 | + x="660" |
| 229 | + y="118" |
| 230 | + width="140" |
| 231 | + height="64" |
| 232 | + rx="8" |
| 233 | + stroke={ok} |
| 234 | + strokeWidth={1.75} |
| 235 | + /> |
| 236 | + <text |
| 237 | + x="730" |
| 238 | + y="143" |
| 239 | + fontSize="12" |
| 240 | + textAnchor="middle" |
| 241 | + fill="currentColor" |
| 242 | + > |
| 243 | + keep order-consistent |
| 244 | + </text> |
| 245 | + <text |
| 246 | + x="730" |
| 247 | + y="164" |
| 248 | + fontSize="13" |
| 249 | + fontWeight={700} |
| 250 | + textAnchor="middle" |
| 251 | + fill={ok} |
| 252 | + > |
| 253 | + κ = 0.66 |
| 254 | + </text> |
| 255 | + |
| 256 | + {/* Proxy branch underneath */} |
| 257 | + <path |
| 258 | + d="M270 190 L270 250" |
| 259 | + stroke="currentColor" |
| 260 | + strokeWidth={1.25} |
| 261 | + opacity={0.4} |
| 262 | + markerEnd="url(#jp-arrow)" |
| 263 | + /> |
| 264 | + <rect |
| 265 | + x="168" |
| 266 | + y="252" |
| 267 | + width="204" |
| 268 | + height="40" |
| 269 | + rx="8" |
| 270 | + stroke="currentColor" |
| 271 | + strokeWidth={1.25} |
| 272 | + strokeDasharray="4 3" |
| 273 | + opacity={0.85} |
| 274 | + /> |
| 275 | + <text |
| 276 | + x="270" |
| 277 | + y="276" |
| 278 | + fontSize="11.5" |
| 279 | + textAnchor="middle" |
| 280 | + fill="currentColor" |
| 281 | + opacity={0.8} |
| 282 | + > |
| 283 | + cheap proxies (geometry, CLIP) |
| 284 | + </text> |
| 285 | + <path |
| 286 | + d="M372 272 C 560 272, 700 272, 724 188" |
| 287 | + stroke="currentColor" |
| 288 | + strokeWidth={1.25} |
| 289 | + opacity={0.4} |
| 290 | + markerEnd="url(#jp-arrow)" |
| 291 | + /> |
| 292 | + <text |
| 293 | + x="540" |
| 294 | + y="305" |
| 295 | + fontSize="10.5" |
| 296 | + textAnchor="middle" |
| 297 | + fill="currentColor" |
| 298 | + opacity={0.5} |
| 299 | + > |
| 300 | + scored against the de-biased judge |
| 301 | + </text> |
| 302 | + |
| 303 | + <defs> |
| 304 | + <marker |
| 305 | + id="jp-arrow" |
| 306 | + viewBox="0 0 10 10" |
| 307 | + refX="8" |
| 308 | + refY="5" |
| 309 | + markerWidth="6" |
| 310 | + markerHeight="6" |
| 311 | + orient="auto-start-reverse" |
| 312 | + > |
| 313 | + <path d="M0 0 L10 5 L0 10 z" fill="currentColor" opacity={0.5} /> |
| 314 | + </marker> |
| 315 | + </defs> |
| 316 | + </svg> |
| 317 | + <figcaption |
| 318 | + style={{ |
| 319 | + fontSize: '0.85rem', |
| 320 | + textAlign: 'center', |
| 321 | + opacity: 0.7, |
| 322 | + marginTop: '0.5rem', |
| 323 | + }} |
| 324 | + > |
| 325 | + The protocol: render each mesh pair from a fixed 24-view rig, ask{' '} |
| 326 | + <em>two independent</em> VLM judges to pick the better one in{' '} |
| 327 | + <em>both</em> presentation orders, and keep only the verdicts that |
| 328 | + survive the swap. Cross-model agreement (Cohen’s κ) is the |
| 329 | + reliability check; the cheap proxies are scored against this same |
| 330 | + reference. |
| 331 | + </figcaption> |
| 332 | + </figure> |
| 333 | + ); |
| 334 | +} |
0 commit comments