copy
Before Width: | Height: | Size: 48 KiB After Width: | Height: | Size: 42 KiB |
1003
assets/swe_bench.svg
Before Width: | Height: | Size: 62 KiB After Width: | Height: | Size: 57 KiB |
Before Width: | Height: | Size: 41 KiB After Width: | Height: | Size: 39 KiB |
|
@ -1,12 +1,12 @@
|
|||
<?xml version="1.0" encoding="utf-8" standalone="no"?>
|
||||
<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
|
||||
"http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
|
||||
<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="720pt" height="432pt" viewBox="0 0 720 432" xmlns="http://www.w3.org/2000/svg" version="1.1">
|
||||
<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="720pt" height="396pt" viewBox="0 0 720 396" xmlns="http://www.w3.org/2000/svg" version="1.1">
|
||||
<metadata>
|
||||
<rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
|
||||
<cc:Work>
|
||||
<dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
|
||||
<dc:date>2024-06-01T07:22:45.471733</dc:date>
|
||||
<dc:date>2024-06-01T07:33:14.155906</dc:date>
|
||||
<dc:format>image/svg+xml</dc:format>
|
||||
<dc:creator>
|
||||
<cc:Agent>
|
||||
|
@ -21,8 +21,8 @@
|
|||
</defs>
|
||||
<g id="figure_1">
|
||||
<g id="patch_1">
|
||||
<path d="M 0 432
|
||||
L 720 432
|
||||
<path d="M 0 396
|
||||
L 720 396
|
||||
L 720 0
|
||||
L 0 0
|
||||
z
|
||||
|
@ -30,8 +30,8 @@ z
|
|||
</g>
|
||||
<g id="axes_1">
|
||||
<g id="patch_2">
|
||||
<path d="M 68.675 343.664
|
||||
L 690 343.664
|
||||
<path d="M 68.675 307.664
|
||||
L 690 307.664
|
||||
L 690 50.4
|
||||
L 68.675 50.4
|
||||
z
|
||||
|
@ -41,17 +41,17 @@ z
|
|||
<g id="xtick_1">
|
||||
<g id="line2d_1">
|
||||
<defs>
|
||||
<path id="md80ddc4930" d="M 0 0
|
||||
<path id="m7e965894c0" d="M 0 0
|
||||
L 0 3.5
|
||||
" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</defs>
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="130.142981" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="130.142981" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_1">
|
||||
<!-- SWE- -->
|
||||
<g style="fill: #555555" transform="translate(109.255481 362.1415) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(109.255481 326.1415) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-53" d="M 894 1481
|
||||
Q 916 1091 1078 847
|
||||
|
@ -135,7 +135,7 @@ z
|
|||
<use xlink:href="#Helvetica-2d" x="227.783203"/>
|
||||
</g>
|
||||
<!-- Agent -->
|
||||
<g style="fill: #555555" transform="translate(109.237981 379.2495) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(109.237981 343.2495) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-41" d="M 2844 1881
|
||||
L 2147 3909
|
||||
|
@ -275,7 +275,7 @@ z
|
|||
<use xlink:href="#Helvetica-74" x="233.544922"/>
|
||||
</g>
|
||||
<!-- + Opus -->
|
||||
<g style="fill: #555555" transform="translate(104.127981 396.9375) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(104.127981 360.9375) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-2b" d="M 288 1369
|
||||
L 288 1894
|
||||
|
@ -412,19 +412,19 @@ z
|
|||
<g id="xtick_2">
|
||||
<g id="line2d_2">
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="213.207821" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="213.207821" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_2">
|
||||
<!-- Open -->
|
||||
<g style="fill: #555555" transform="translate(193.639071 362.454) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(193.639071 326.454) scale(0.16 -0.16)">
|
||||
<use xlink:href="#Helvetica-4f"/>
|
||||
<use xlink:href="#Helvetica-70" x="77.783203"/>
|
||||
<use xlink:href="#Helvetica-65" x="133.398438"/>
|
||||
<use xlink:href="#Helvetica-6e" x="189.013672"/>
|
||||
</g>
|
||||
<!-- Devin -->
|
||||
<g style="fill: #555555" transform="translate(192.755321 379.562) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(192.755321 343.562) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-44" d="M 2250 531
|
||||
Q 2566 531 2769 597
|
||||
|
@ -482,19 +482,19 @@ z
|
|||
<g id="xtick_3">
|
||||
<g id="line2d_3">
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="296.27266" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="296.27266" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_3">
|
||||
<!-- SWE- -->
|
||||
<g style="fill: #555555" transform="translate(275.38516 362.1415) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(275.38516 326.1415) scale(0.16 -0.16)">
|
||||
<use xlink:href="#Helvetica-53"/>
|
||||
<use xlink:href="#Helvetica-57" x="66.699219"/>
|
||||
<use xlink:href="#Helvetica-45" x="161.083984"/>
|
||||
<use xlink:href="#Helvetica-2d" x="227.783203"/>
|
||||
</g>
|
||||
<!-- Agent -->
|
||||
<g style="fill: #555555" transform="translate(275.36766 379.2495) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(275.36766 343.2495) scale(0.16 -0.16)">
|
||||
<use xlink:href="#Helvetica-41"/>
|
||||
<use xlink:href="#Helvetica-67" x="66.699219"/>
|
||||
<use xlink:href="#Helvetica-65" x="122.314453"/>
|
||||
|
@ -502,7 +502,7 @@ z
|
|||
<use xlink:href="#Helvetica-74" x="233.544922"/>
|
||||
</g>
|
||||
<!-- + GPT-4 -->
|
||||
<g style="fill: #555555" transform="translate(265.82016 396.5625) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(265.82016 360.5625) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-47" d="M 2472 4709
|
||||
Q 3119 4709 3591 4459
|
||||
|
@ -598,12 +598,12 @@ z
|
|||
<g id="xtick_4">
|
||||
<g id="line2d_4">
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="379.3375" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="379.3375" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_4">
|
||||
<!-- AutoCode -->
|
||||
<g style="fill: #555555" transform="translate(343.7575 362.1415) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(343.7575 326.1415) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-6f" d="M 1741 363
|
||||
Q 2300 363 2508 786
|
||||
|
@ -690,7 +690,7 @@ z
|
|||
<use xlink:href="#Helvetica-65" x="389.160156"/>
|
||||
</g>
|
||||
<!-- Rover -->
|
||||
<g style="fill: #555555" transform="translate(357.99875 379.2495) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(357.99875 343.2495) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-52" d="M 2622 2488
|
||||
Q 3059 2488 3314 2663
|
||||
|
@ -754,12 +754,12 @@ z
|
|||
<g id="xtick_5">
|
||||
<g id="line2d_5">
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="462.40234" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="462.40234" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_5">
|
||||
<!-- Amazon Q -->
|
||||
<g style="fill: #555555" transform="translate(424.61109 362.1415) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(424.61109 326.1415) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-6d" d="M 413 3347
|
||||
L 969 3347
|
||||
|
@ -895,7 +895,7 @@ z
|
|||
<use xlink:href="#Helvetica-51" x="394.628906"/>
|
||||
</g>
|
||||
<!-- Developer -->
|
||||
<g style="fill: #555555" transform="translate(425.93984 379.2495) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(425.93984 343.2495) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-6c" d="M 428 4591
|
||||
L 991 4591
|
||||
|
@ -916,7 +916,7 @@ z
|
|||
<use xlink:href="#Helvetica-72" x="422.509766"/>
|
||||
</g>
|
||||
<!-- Agent -->
|
||||
<g style="fill: #555555" transform="translate(441.49734 396.3575) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(441.49734 360.3575) scale(0.16 -0.16)">
|
||||
<use xlink:href="#Helvetica-41"/>
|
||||
<use xlink:href="#Helvetica-67" x="66.699219"/>
|
||||
<use xlink:href="#Helvetica-65" x="122.314453"/>
|
||||
|
@ -928,12 +928,12 @@ z
|
|||
<g id="xtick_6">
|
||||
<g id="line2d_6">
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="545.467179" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="545.467179" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_6">
|
||||
<!-- Aider -->
|
||||
<g style="fill: #555555" transform="translate(521.435929 362.8215) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(521.435929 326.8215) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="DejaVuSans-Bold-41" d="M 3419 850
|
||||
L 1538 850
|
||||
|
@ -1040,7 +1040,7 @@ z
|
|||
<use xlink:href="#DejaVuSans-Bold-72" x="251.074219"/>
|
||||
</g>
|
||||
<!-- GPT-4o -->
|
||||
<g style="fill: #555555" transform="translate(514.377179 380.738) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(514.377179 344.738) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="DejaVuSans-Bold-47" d="M 4781 347
|
||||
Q 4331 128 3847 18
|
||||
|
@ -1159,12 +1159,12 @@ z
|
|||
<g id="xtick_7">
|
||||
<g id="line2d_7">
|
||||
<g>
|
||||
<use xlink:href="#md80ddc4930" x="628.532019" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m7e965894c0" x="628.532019" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_7">
|
||||
<!-- Aider -->
|
||||
<g style="fill: #555555" transform="translate(604.500769 362.8215) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(604.500769 326.8215) scale(0.16 -0.16)">
|
||||
<use xlink:href="#DejaVuSans-Bold-41"/>
|
||||
<use xlink:href="#DejaVuSans-Bold-69" x="77.392578"/>
|
||||
<use xlink:href="#DejaVuSans-Bold-64" x="111.669922"/>
|
||||
|
@ -1172,7 +1172,7 @@ z
|
|||
<use xlink:href="#DejaVuSans-Bold-72" x="251.074219"/>
|
||||
</g>
|
||||
<!-- GPT-4o -->
|
||||
<g style="fill: #555555" transform="translate(597.442019 380.738) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(597.442019 344.738) scale(0.16 -0.16)">
|
||||
<use xlink:href="#DejaVuSans-Bold-47"/>
|
||||
<use xlink:href="#DejaVuSans-Bold-50" x="82.080078"/>
|
||||
<use xlink:href="#DejaVuSans-Bold-54" x="155.371094"/>
|
||||
|
@ -1181,7 +1181,7 @@ z
|
|||
<use xlink:href="#DejaVuSans-Bold-6f" x="319.917969"/>
|
||||
</g>
|
||||
<!-- & Opus -->
|
||||
<g style="fill: #555555" transform="translate(595.787019 398.6545) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(595.787019 362.6545) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="DejaVuSans-Bold-26" d="M 2497 3097
|
||||
L 3775 1691
|
||||
|
@ -1339,23 +1339,23 @@ z
|
|||
<g id="matplotlib.axis_2">
|
||||
<g id="ytick_1">
|
||||
<g id="line2d_8">
|
||||
<path d="M 68.675 343.664
|
||||
L 690 343.664
|
||||
" clip-path="url(#p837c6e1836)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
<path d="M 68.675 307.664
|
||||
L 690 307.664
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="line2d_9">
|
||||
<defs>
|
||||
<path id="ma95ca84bab" d="M 0 0
|
||||
<path id="m46e428bacd" d="M 0 0
|
||||
L -3.5 0
|
||||
" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</defs>
|
||||
<g>
|
||||
<use xlink:href="#ma95ca84bab" x="68.675" y="343.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m46e428bacd" x="68.675" y="307.664" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_8">
|
||||
<!-- 0 -->
|
||||
<g transform="translate(56.114063 347.250719) scale(0.1 -0.1)">
|
||||
<g transform="translate(56.114063 311.250719) scale(0.1 -0.1)">
|
||||
<defs>
|
||||
<path id="Helvetica-30" d="M 1731 4475
|
||||
Q 2600 4475 2988 3759
|
||||
|
@ -1385,18 +1385,18 @@ z
|
|||
</g>
|
||||
<g id="ytick_2">
|
||||
<g id="line2d_10">
|
||||
<path d="M 68.675 290.565322
|
||||
L 690 290.565322
|
||||
" clip-path="url(#p837c6e1836)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
<path d="M 68.675 261.083518
|
||||
L 690 261.083518
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="line2d_11">
|
||||
<g>
|
||||
<use xlink:href="#ma95ca84bab" x="68.675" y="290.565322" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m46e428bacd" x="68.675" y="261.083518" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_9">
|
||||
<!-- 5 -->
|
||||
<g transform="translate(56.114063 294.15204) scale(0.1 -0.1)">
|
||||
<g transform="translate(56.114063 264.670237) scale(0.1 -0.1)">
|
||||
<defs>
|
||||
<path id="Helvetica-35" d="M 791 1141
|
||||
Q 847 659 1238 475
|
||||
|
@ -1431,18 +1431,18 @@ z
|
|||
</g>
|
||||
<g id="ytick_3">
|
||||
<g id="line2d_12">
|
||||
<path d="M 68.675 237.466643
|
||||
L 690 237.466643
|
||||
" clip-path="url(#p837c6e1836)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
<path d="M 68.675 214.503037
|
||||
L 690 214.503037
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="line2d_13">
|
||||
<g>
|
||||
<use xlink:href="#ma95ca84bab" x="68.675" y="237.466643" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m46e428bacd" x="68.675" y="214.503037" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_10">
|
||||
<!-- 10 -->
|
||||
<g transform="translate(50.553125 241.053362) scale(0.1 -0.1)">
|
||||
<g transform="translate(50.553125 218.089756) scale(0.1 -0.1)">
|
||||
<defs>
|
||||
<path id="Helvetica-31" d="M 613 3169
|
||||
L 613 3600
|
||||
|
@ -1463,18 +1463,18 @@ z
|
|||
</g>
|
||||
<g id="ytick_4">
|
||||
<g id="line2d_14">
|
||||
<path d="M 68.675 184.367965
|
||||
L 690 184.367965
|
||||
" clip-path="url(#p837c6e1836)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
<path d="M 68.675 167.922555
|
||||
L 690 167.922555
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="line2d_15">
|
||||
<g>
|
||||
<use xlink:href="#ma95ca84bab" x="68.675" y="184.367965" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m46e428bacd" x="68.675" y="167.922555" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_11">
|
||||
<!-- 15 -->
|
||||
<g transform="translate(50.553125 187.954684) scale(0.1 -0.1)">
|
||||
<g transform="translate(50.553125 171.509274) scale(0.1 -0.1)">
|
||||
<use xlink:href="#Helvetica-31"/>
|
||||
<use xlink:href="#Helvetica-35" x="55.615234"/>
|
||||
</g>
|
||||
|
@ -1482,18 +1482,18 @@ L 690 184.367965
|
|||
</g>
|
||||
<g id="ytick_5">
|
||||
<g id="line2d_16">
|
||||
<path d="M 68.675 131.269287
|
||||
L 690 131.269287
|
||||
" clip-path="url(#p837c6e1836)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
<path d="M 68.675 121.342074
|
||||
L 690 121.342074
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="line2d_17">
|
||||
<g>
|
||||
<use xlink:href="#ma95ca84bab" x="68.675" y="131.269287" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m46e428bacd" x="68.675" y="121.342074" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_12">
|
||||
<!-- 20 -->
|
||||
<g transform="translate(50.553125 134.856006) scale(0.1 -0.1)">
|
||||
<g transform="translate(50.553125 124.928792) scale(0.1 -0.1)">
|
||||
<defs>
|
||||
<path id="Helvetica-32" d="M 200 0
|
||||
Q 231 578 439 1006
|
||||
|
@ -1528,18 +1528,18 @@ z
|
|||
</g>
|
||||
<g id="ytick_6">
|
||||
<g id="line2d_18">
|
||||
<path d="M 68.675 78.170609
|
||||
L 690 78.170609
|
||||
" clip-path="url(#p837c6e1836)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
<path d="M 68.675 74.761592
|
||||
L 690 74.761592
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: none; stroke: #b0b0b0; stroke-width: 0.2; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="line2d_19">
|
||||
<g>
|
||||
<use xlink:href="#ma95ca84bab" x="68.675" y="78.170609" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
<use xlink:href="#m46e428bacd" x="68.675" y="74.761592" style="stroke: #000000; stroke-width: 0.8"/>
|
||||
</g>
|
||||
</g>
|
||||
<g id="text_13">
|
||||
<!-- 25 -->
|
||||
<g transform="translate(50.553125 81.757327) scale(0.1 -0.1)">
|
||||
<g transform="translate(50.553125 78.348311) scale(0.1 -0.1)">
|
||||
<use xlink:href="#Helvetica-32"/>
|
||||
<use xlink:href="#Helvetica-35" x="55.615234"/>
|
||||
</g>
|
||||
|
@ -1547,7 +1547,7 @@ L 690 78.170609
|
|||
</g>
|
||||
<g id="text_14">
|
||||
<!-- Instances resolved (%) -->
|
||||
<g style="fill: #555555" transform="translate(42.80125 288.564812) rotate(-90) scale(0.18 -0.18)">
|
||||
<g style="fill: #555555" transform="translate(42.80125 270.564812) rotate(-90) scale(0.18 -0.18)">
|
||||
<defs>
|
||||
<path id="Helvetica-49" d="M 628 4591
|
||||
L 1256 4591
|
||||
|
@ -1684,18 +1684,18 @@ z
|
|||
</g>
|
||||
</g>
|
||||
<g id="patch_3">
|
||||
<path d="M 68.675 343.664
|
||||
<path d="M 68.675 307.664
|
||||
L 68.675 50.4
|
||||
" style="fill: none; stroke: #dddddd; stroke-width: 0.5; stroke-linejoin: miter; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="patch_4">
|
||||
<path d="M 690 343.664
|
||||
<path d="M 690 307.664
|
||||
L 690 50.4
|
||||
" style="fill: none; stroke: #dddddd; stroke-width: 0.5; stroke-linejoin: miter; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="patch_5">
|
||||
<path d="M 68.675 343.664
|
||||
L 690 343.664
|
||||
<path d="M 68.675 307.664
|
||||
L 690 307.664
|
||||
" style="fill: none; stroke: #dddddd; stroke-width: 0.5; stroke-linejoin: miter; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="patch_6">
|
||||
|
@ -1704,64 +1704,64 @@ L 690 50.4
|
|||
" style="fill: none; stroke: #dddddd; stroke-width: 0.5; stroke-linejoin: miter; stroke-linecap: square"/>
|
||||
</g>
|
||||
<g id="patch_7">
|
||||
<path d="M 96.917045 343.664
|
||||
L 163.368917 343.664
|
||||
L 163.368917 219.413093
|
||||
L 96.917045 219.413093
|
||||
<path d="M 96.917045 307.664
|
||||
L 163.368917 307.664
|
||||
L 163.368917 198.665673
|
||||
L 96.917045 198.665673
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
</g>
|
||||
<g id="patch_8">
|
||||
<path d="M 179.981885 343.664
|
||||
L 246.433757 343.664
|
||||
L 246.433757 166.314415
|
||||
L 179.981885 166.314415
|
||||
<path d="M 179.981885 307.664
|
||||
L 246.433757 307.664
|
||||
L 246.433757 152.085191
|
||||
L 179.981885 152.085191
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
</g>
|
||||
<g id="patch_9">
|
||||
<path d="M 263.046725 343.664
|
||||
L 329.498596 343.664
|
||||
L 329.498596 152.508758
|
||||
L 263.046725 152.508758
|
||||
<path d="M 263.046725 307.664
|
||||
L 329.498596 307.664
|
||||
L 329.498596 139.974266
|
||||
L 263.046725 139.974266
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
</g>
|
||||
<g id="patch_10">
|
||||
<path d="M 346.111564 343.664
|
||||
L 412.563436 343.664
|
||||
L 412.563436 141.889023
|
||||
L 346.111564 141.889023
|
||||
<path d="M 346.111564 307.664
|
||||
L 412.563436 307.664
|
||||
L 412.563436 130.65817
|
||||
L 346.111564 130.65817
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
</g>
|
||||
<g id="patch_11">
|
||||
<path d="M 429.176404 343.664
|
||||
L 495.628275 343.664
|
||||
L 495.628275 128.083366
|
||||
L 429.176404 128.083366
|
||||
<path d="M 429.176404 307.664
|
||||
L 495.628275 307.664
|
||||
L 495.628275 118.547245
|
||||
L 429.176404 118.547245
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #b3d1e6; opacity: 0.3"/>
|
||||
</g>
|
||||
<g id="patch_12">
|
||||
<path d="M 512.241243 343.664
|
||||
L 578.693115 343.664
|
||||
L 578.693115 78.170609
|
||||
L 512.241243 78.170609
|
||||
<path d="M 512.241243 307.664
|
||||
L 578.693115 307.664
|
||||
L 578.693115 74.761592
|
||||
L 512.241243 74.761592
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #17965a; opacity: 0.9"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #17965a; opacity: 0.9"/>
|
||||
</g>
|
||||
<g id="patch_13">
|
||||
<path d="M 595.306083 343.664
|
||||
L 661.757955 343.664
|
||||
L 661.757955 64.364952
|
||||
L 595.306083 64.364952
|
||||
<path d="M 595.306083 307.664
|
||||
L 661.757955 307.664
|
||||
L 661.757955 62.650667
|
||||
L 595.306083 62.650667
|
||||
z
|
||||
" clip-path="url(#p837c6e1836)" style="fill: #17965a; opacity: 0.9"/>
|
||||
" clip-path="url(#p64bcd2b177)" style="fill: #17965a; opacity: 0.9"/>
|
||||
</g>
|
||||
<g id="text_15">
|
||||
<!-- 11.7% -->
|
||||
<g style="fill: #555555" transform="translate(107.460481 241.510329) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(107.460481 219.459269) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-2e" d="M 547 681
|
||||
L 1200 681
|
||||
|
@ -1794,7 +1794,7 @@ z
|
|||
</g>
|
||||
<g id="text_16">
|
||||
<!-- 16.7% -->
|
||||
<g style="fill: #555555" transform="translate(190.525321 188.41165) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(190.525321 172.878788) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-36" d="M 1872 4494
|
||||
Q 2622 4494 2917 4105
|
||||
|
@ -1836,7 +1836,7 @@ z
|
|||
</g>
|
||||
<g id="text_17">
|
||||
<!-- 18.0% -->
|
||||
<g style="fill: #555555" transform="translate(273.59016 174.605994) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(273.59016 160.767862) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-38" d="M 1741 2600
|
||||
Q 2113 2600 2322 2808
|
||||
|
@ -1887,7 +1887,7 @@ z
|
|||
</g>
|
||||
<g id="text_18">
|
||||
<!-- 19.0% -->
|
||||
<g style="fill: #555555" transform="translate(356.655 163.986258) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(356.655 151.451766) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-39" d="M 850 1081
|
||||
Q 875 616 1209 438
|
||||
|
@ -1929,7 +1929,7 @@ z
|
|||
</g>
|
||||
<g id="text_19">
|
||||
<!-- 20.3% -->
|
||||
<g style="fill: #555555" transform="translate(439.71984 150.180602) scale(0.16 -0.16)">
|
||||
<g style="fill: #555555" transform="translate(439.71984 139.340841) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="Helvetica-33" d="M 1663 -122
|
||||
Q 869 -122 511 314
|
||||
|
@ -1975,7 +1975,7 @@ z
|
|||
</g>
|
||||
<g id="text_20">
|
||||
<!-- 25.0% -->
|
||||
<g style="fill: #eeeeee" transform="translate(517.713429 100.947844) scale(0.16 -0.16)">
|
||||
<g style="fill: #eeeeee" transform="translate(517.713429 96.235188) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="DejaVuSans-Bold-32" d="M 1844 884
|
||||
L 3897 884
|
||||
|
@ -2109,7 +2109,7 @@ z
|
|||
</g>
|
||||
<g id="text_21">
|
||||
<!-- 26.3% -->
|
||||
<g style="fill: #eeeeee" transform="translate(600.778269 87.142188) scale(0.16 -0.16)">
|
||||
<g style="fill: #eeeeee" transform="translate(600.778269 84.124263) scale(0.16 -0.16)">
|
||||
<defs>
|
||||
<path id="DejaVuSans-Bold-36" d="M 2316 2303
|
||||
Q 2000 2303 1842 2098
|
||||
|
@ -2265,8 +2265,8 @@ z
|
|||
</g>
|
||||
</g>
|
||||
<defs>
|
||||
<clipPath id="p837c6e1836">
|
||||
<rect x="68.675" y="50.4" width="621.325" height="293.264"/>
|
||||
<clipPath id="p64bcd2b177">
|
||||
<rect x="68.675" y="50.4" width="621.325" height="257.264"/>
|
||||
</clipPath>
|
||||
</defs>
|
||||
</svg>
|
||||
|
|
Before Width: | Height: | Size: 53 KiB After Width: | Height: | Size: 53 KiB |
|
@ -14,15 +14,24 @@ def plot_swe_bench(data_file, is_lite):
|
|||
|
||||
models = []
|
||||
pass_rates = []
|
||||
|
||||
instances = []
|
||||
for line in lines:
|
||||
if line.strip():
|
||||
pass_rate, model = line.split("%")
|
||||
model = model.strip()
|
||||
if "(" in model:
|
||||
pieces = model.split("(")
|
||||
model = pieces[0]
|
||||
ins = pieces[1].strip(")")
|
||||
else:
|
||||
ins = None
|
||||
instances.insert(0, ins)
|
||||
model = model.replace("|", "\n")
|
||||
models.insert(0, model.strip())
|
||||
pass_rates.insert(0, float(pass_rate.strip()))
|
||||
|
||||
dump(instances)
|
||||
|
||||
plt.rcParams["hatch.linewidth"] = 0.5
|
||||
plt.rcParams["hatch.color"] = "#444444"
|
||||
|
||||
|
@ -36,7 +45,7 @@ def plot_swe_bench(data_file, is_lite):
|
|||
rc("font", **font_params)
|
||||
plt.rcParams["text.color"] = font_color
|
||||
|
||||
fig, ax = plt.subplots(figsize=(10, 6))
|
||||
fig, ax = plt.subplots(figsize=(10, 5.5))
|
||||
ax.grid(axis="y", zorder=0, lw=0.2)
|
||||
for spine in ax.spines.values():
|
||||
spine.set_edgecolor("#DDDDDD")
|
||||
|
@ -73,6 +82,23 @@ def plot_swe_bench(data_file, is_lite):
|
|||
fontfamily=fontfamily,
|
||||
)
|
||||
|
||||
for model, ins, bar in zip(models, instances, bars):
|
||||
if not ins:
|
||||
continue
|
||||
yval = bar.get_height()
|
||||
y = yval - 2.5
|
||||
va = "top"
|
||||
color = "#eee" if "Aider" in model else "#555"
|
||||
ax.text(
|
||||
bar.get_x() + bar.get_width() / 2,
|
||||
y,
|
||||
f"of {ins}",
|
||||
ha="center",
|
||||
va=va,
|
||||
fontsize=12,
|
||||
color=color,
|
||||
)
|
||||
|
||||
# ax.set_xlabel("Models", fontsize=18)
|
||||
ax.set_ylabel("Instances resolved (%)", fontsize=18, color=font_color)
|
||||
if is_lite:
|
||||
|
@ -86,22 +112,7 @@ def plot_swe_bench(data_file, is_lite):
|
|||
color=font_color,
|
||||
)
|
||||
|
||||
if is_lite:
|
||||
plt.tight_layout(pad=3.0)
|
||||
else:
|
||||
# Add note at the bottom of the graph
|
||||
note = "(570) and (2294) denote the number of SWE Bench instances benchmarked"
|
||||
plt.figtext(
|
||||
0.5,
|
||||
0.05,
|
||||
note,
|
||||
wrap=True,
|
||||
horizontalalignment="center",
|
||||
fontsize=12,
|
||||
color=font_color,
|
||||
)
|
||||
|
||||
plt.tight_layout(pad=3.0, rect=[0, 0.05, 1, 1])
|
||||
plt.tight_layout(pad=3.0)
|
||||
|
||||
out_fname = Path(data_file.replace("-", "_"))
|
||||
plt.savefig(out_fname.with_suffix(".jpg").name)
|
||||
|
|