form-field-vlm-v2 / assets /headline-e2e.svg
hung-k-nguyen's picture
form-field-vlm-v2 scorecard + leaderboard (GPT-6 Astra included)
2112894
|
Raw History Blame Contribute Delete
3.5 kB
<svg xmlns="http://www.w3.org/2000/svg" width="720" height="448" viewBox="0 0 720 448" font-family="system-ui,-apple-system,'Segoe UI',sans-serif" fill="#18232d">
<rect width="100%" height="100%" fill="#fff"/>
<text x="24" y="28" font-size="19" font-weight="700">End-to-end F1 — form-field extraction</text>
<text x="24" y="48" font-size="12.5" fill="#5d6b77">354-page benchmark · box + type + value all correct · IoU ≥ 0.5</text>
<text x="200" y="78" font-size="12.5" text-anchor="end">Nutrient Hybrid (detector + VLM)</text>
<rect x="210" y="67" width="380.0" height="16" rx="2" fill="#0f5a48"/>
<text x="598.0" y="79" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.565</text>
<text x="200" y="108" font-size="12.5" text-anchor="end">Nutrient VLM v2</text>
<rect x="210" y="97" width="348.4" height="16" rx="2" fill="#16705a"/>
<text x="566.4" y="109" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.518</text>
<text x="200" y="138" font-size="12.5" text-anchor="end">GPT-6 Astra</text>
<rect x="210" y="127" width="279.3" height="16" rx="2" fill="#9aa7b2"/>
<text x="497.3" y="139" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.415</text>
<text x="200" y="168" font-size="12.5" text-anchor="end">GPT-5.6 Sol</text>
<rect x="210" y="157" width="227.5" height="16" rx="2" fill="#9aa7b2"/>
<text x="445.5" y="169" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.338</text>
<text x="200" y="198" font-size="12.5" text-anchor="end">GPT-5.6 Terra</text>
<rect x="210" y="187" width="212.0" height="16" rx="2" fill="#9aa7b2"/>
<text x="430.0" y="199" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.315</text>
<text x="200" y="228" font-size="12.5" text-anchor="end">Gemini Flash</text>
<rect x="210" y="217" width="164.9" height="16" rx="2" fill="#9aa7b2"/>
<text x="382.9" y="229" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.245</text>
<text x="200" y="258" font-size="12.5" text-anchor="end">Gemini Pro</text>
<rect x="210" y="247" width="145.3" height="16" rx="2" fill="#9aa7b2"/>
<text x="363.3" y="259" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.216</text>
<text x="200" y="288" font-size="12.5" text-anchor="end">Gemini Flash-Lite</text>
<rect x="210" y="277" width="112.8" height="16" rx="2" fill="#9aa7b2"/>
<text x="330.8" y="289" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.168</text>
<text x="200" y="318" font-size="12.5" text-anchor="end">GPT-5.6 Luna</text>
<rect x="210" y="307" width="89.9" height="16" rx="2" fill="#9aa7b2"/>
<text x="307.9" y="319" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.134</text>
<text x="200" y="348" font-size="12.5" text-anchor="end">Claude Haiku 4.5</text>
<rect x="210" y="337" width="7.3" height="16" rx="2" fill="#9aa7b2"/>
<text x="225.3" y="349" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.011</text>
<text x="200" y="378" font-size="12.5" text-anchor="end">Claude Opus 4.8</text>
<rect x="210" y="367" width="7.0" height="16" rx="2" fill="#9aa7b2"/>
<text x="225.0" y="379" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.010</text>
<text x="200" y="408" font-size="12.5" text-anchor="end">Claude Sonnet 5</text>
<rect x="210" y="397" width="6.0" height="16" rx="2" fill="#9aa7b2"/>
<text x="224.0" y="409" font-size="12" font-family="ui-monospace,monospace" fill="#5d6b77">0.009</text>
</svg>