Spaces:

AryaWu
/

tokenmaxxing_trajectory_viewer

Running

App Files Files Community

tokenmaxxing_trajectory_viewer / graphs /all_transitions_python.html

Aryarya

build

4b4d08b about 18 hours ago

Raw

History Blame Contribute Delete

173 kB

	<!DOCTYPE html>
	<html lang="en">
	<head>
	<meta charset="UTF-8">
	<meta name="viewport" content="width=device-width,initial-scale=1">
	<title>Transition Probabilities — Merged Across Problems</title>
	<script src="https://unpkg.com/cytoscape@3.29.2/dist/cytoscape.min.js"></script>
	<style>
	* { box-sizing: border-box; margin: 0; padding: 0; }
	body { font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
	background: #f8fafc; color: #1e293b; font-size: 13px; }
	h1 { font-size: 1.3rem; font-weight: 700; }
	.page-header { padding: 16px 28px 10px; border-bottom: 1px solid #e2e8f0;
	background: white; display: flex; align-items: baseline; gap: 14px; flex-wrap: wrap; }
	.subtitle { font-size: 0.78rem; color: #64748b; }

	.main { display: flex; gap: 12px; align-items: flex-start; padding: 14px 28px; }
	.cy-wrap { flex: 1 1 auto; min-width: 0; border: 1px solid #e2e8f0; border-radius: 8px;
	background: white; overflow: hidden; }
	#cy { width: 100%; height: 700px; }
	.detail-panel { flex: 0 0 360px; border: 1px solid #e2e8f0; border-radius: 8px;
	background: white; padding: 12px; font-size: 0.75rem;
	height: 700px; overflow-y: auto; }
	.detail-placeholder { color: #94a3b8; font-style: italic; padding: 8px 0; }

	.det-title { font-weight: 700; margin-bottom: 8px; font-size: 0.9rem;
	display: flex; align-items: center; gap: 6px; flex-wrap: wrap; }
	.det-def { font-size: 0.72rem; color: #475569; margin-bottom: 8px; line-height: 1.45; }
	.det-row { display: flex; gap: 6px; margin-bottom: 3px; align-items: baseline; }
	.det-lbl { color: #64748b; min-width: 96px; flex-shrink: 0; font-size: 0.72rem; }
	.det-val { color: #1e293b; font-family: monospace; font-size: 0.78rem; }
	.state-chip { font-size: 0.74rem; font-weight: 700; color: white; padding: 2px 8px;
	border-radius: 4px; white-space: nowrap; }
	.det-sub { font-size: 0.68rem; font-weight: 700; color: #64748b; margin: 10px 0 5px;
	text-transform: uppercase; }

	/* breakdown bars */
	.bd-row { margin-bottom: 10px; }
	.bd-head { display: flex; justify-content: space-between; font-size: 0.72rem;
	gap: 6px; margin-bottom: 3px; align-items: flex-start; }
	.bd-lbl { font-weight: 600; flex: 1 1 auto; min-width: 0;
	display: flex; flex-wrap: wrap; gap: 2px; align-items: center; }
	.bd-num { font-family: monospace; color: #475569; flex-shrink: 0; }
	.bd-bar { height: 8px; border-radius: 4px; background: #eef2f7; overflow: hidden; }
	.bd-fill { height: 100%; border-radius: 4px; }
	.bd-ex { font-size: 0.64rem; font-family: ui-monospace, monospace; color: #64748b;
	padding: 2px 0 2px 6px; border-left: 2px solid #eef2f7; margin-top: 2px;
	white-space: nowrap; overflow: hidden; text-overflow: ellipsis; }
	.kind-pill { display: inline-block; font-size: 0.62rem; padding: 1px 6px; border-radius: 3px;
	font-family: ui-monospace, monospace; font-weight: 600; line-height: 1.4; }
	.kind-arrow { display: inline-block; color: #94a3b8; font-size: 0.7rem; margin: 0 1px; }

	/* gap label distribution bars */
	.lbl-row { margin-bottom: 7px; }
	.lbl-head { display: flex; justify-content: space-between; font-size: 0.72rem;
	gap: 6px; margin-bottom: 3px; align-items: center; }
	.lbl-pill { display: inline-block; font-size: 0.65rem; padding: 1px 7px; border-radius: 10px;
	font-weight: 700; white-space: nowrap; }
	.lbl-num { font-family: monospace; color: #475569; flex-shrink: 0; font-size: 0.7rem; }
	.lbl-bar { height: 8px; border-radius: 4px; background: #eef2f7; overflow: hidden; }
	.lbl-fill { height: 100%; border-radius: 4px; }

	/* outgoing distribution list (node detail) */
	.out-row { display: flex; align-items: center; gap: 6px; margin-bottom: 5px; }
	.out-tgt { width: 96px; flex-shrink: 0; font-size: 0.72rem; font-weight: 600; }
	.out-bar { flex: 1 1 auto; height: 10px; background: #eef2f7; border-radius: 5px; overflow: hidden; }
	.out-fill{ height: 100%; border-radius: 5px; }
	.out-num { width: 78px; flex-shrink: 0; text-align: right; font-family: monospace;
	font-size: 0.72rem; color: #475569; }

	/* NLP metric distribution bars */
	.nlp-row { display: flex; align-items: center; gap: 6px; margin-bottom: 4px; }
	.nlp-lbl { width: 140px; flex-shrink: 0; font-size: 0.7rem; color: #374151; }
	.nlp-bar { flex: 1 1 auto; height: 8px; background: #eef2f7; border-radius: 4px; overflow: hidden; }
	.nlp-fill{ height: 100%; border-radius: 4px; background: #6366f1; }
	.nlp-num { width: 44px; flex-shrink: 0; text-align: right; font-family: monospace;
	font-size: 0.7rem; color: #475569; }
	.nlp-chip { display: inline-block; font-size: 0.64rem; padding: 1px 6px; border-radius: 10px;
	font-weight: 700; background: #eef2ff; color: #4338ca; margin: 1px; white-space: nowrap; }

	.table-section { padding: 6px 28px 32px; }
	.table-section h2 { font-size: 0.95rem; margin-bottom: 8px; }
	table.verify { border-collapse: collapse; font-size: 0.74rem; }
	table.verify th, table.verify td { border: 1px solid #e2e8f0; padding: 4px 9px; text-align: center; }
	table.verify thead th { background: #f1f5f9; }
	table.verify td.tr-name { text-align: left; font-weight: 700; font-family: monospace; }
	table.verify td.tr-total { font-weight: 700; background: #f8fafc; font-family: monospace; }
	table.verify td.tr-sum { font-weight: 700; color: #16a34a; font-family: monospace; }
	</style>
	</head>
	<body>
	<div class="page-header">
	<h1>Transition Probabilities</h1>
	<span class="subtitle">python · all models · 381 problems merged · edge label = count · click a node or edge for details</span>
	</div>

	<div class="main">
	<div class="cy-wrap"><div id="cy"></div></div>
	<div class="detail-panel" id="detail">
	<div class="detail-placeholder">Click a node or edge for details</div>
	</div>
	</div>

	<div class="table-section">
	<h2>Verification — outgoing probability per source state (sums to 100%)</h2>
	<table class="verify">
	<thead><tr><th class="tr-name" style="text-align:left">source</th><th>out total</th><th>Σ%</th>
	<th style="text-align:left">→ targets</th></tr></thead>
	<tbody><tr><td class="tr-name">start</td><td class="tr-total">381</td><td class="tr-sum">100%</td><td style="text-align:left">breakthrough <b>82.2%</b> <span style="color:#94a3b8">(313)</span>   stuck <b>8.9%</b> <span style="color:#94a3b8">(34)</span>   improved <b>8.9%</b> <span style="color:#94a3b8">(34)</span></td></tr><tr><td class="tr-name">improved</td><td class="tr-total">84</td><td class="tr-sum">100%</td><td style="text-align:left">stuck <b>28.6%</b> <span style="color:#94a3b8">(24)</span>   breakthrough <b>19.0%</b> <span style="color:#94a3b8">(16)</span>   regressed <b>16.7%</b> <span style="color:#94a3b8">(14)</span>   end_limit <b>15.5%</b> <span style="color:#94a3b8">(13)</span>   improved <b>7.1%</b> <span style="color:#94a3b8">(6)</span>   broke <b>6.0%</b> <span style="color:#94a3b8">(5)</span>   churn <b>6.0%</b> <span style="color:#94a3b8">(5)</span>   end_submit <b>1.2%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">churn</td><td class="tr-total">25</td><td class="tr-sum">100%</td><td style="text-align:left">churn <b>44.0%</b> <span style="color:#94a3b8">(11)</span>   stuck <b>20.0%</b> <span style="color:#94a3b8">(5)</span>   end_limit <b>16.0%</b> <span style="color:#94a3b8">(4)</span>   broke <b>8.0%</b> <span style="color:#94a3b8">(2)</span>   breakthrough <b>8.0%</b> <span style="color:#94a3b8">(2)</span>   improved <b>4.0%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">stuck</td><td class="tr-total">153</td><td class="tr-sum">100%</td><td style="text-align:left">stuck <b>47.1%</b> <span style="color:#94a3b8">(72)</span>   improved <b>18.3%</b> <span style="color:#94a3b8">(28)</span>   breakthrough <b>13.7%</b> <span style="color:#94a3b8">(21)</span>   end_limit <b>7.8%</b> <span style="color:#94a3b8">(12)</span>   end_submit <b>4.6%</b> <span style="color:#94a3b8">(7)</span>   broke <b>3.9%</b> <span style="color:#94a3b8">(6)</span>   churn <b>2.6%</b> <span style="color:#94a3b8">(4)</span>   regressed <b>2.0%</b> <span style="color:#94a3b8">(3)</span></td></tr><tr><td class="tr-name">breakthrough</td><td class="tr-total">362</td><td class="tr-sum">100%</td><td style="text-align:left">end_submit <b>79.0%</b> <span style="color:#94a3b8">(286)</span>   stay <b>18.8%</b> <span style="color:#94a3b8">(68)</span>   regressed <b>1.7%</b> <span style="color:#94a3b8">(6)</span>   broke <b>0.6%</b> <span style="color:#94a3b8">(2)</span></td></tr><tr><td class="tr-name">stay</td><td class="tr-total">302</td><td class="tr-sum">100%</td><td style="text-align:left">stay <b>77.5%</b> <span style="color:#94a3b8">(234)</span>   end_submit <b>14.9%</b> <span style="color:#94a3b8">(45)</span>   end_limit <b>3.6%</b> <span style="color:#94a3b8">(11)</span>   broke <b>3.0%</b> <span style="color:#94a3b8">(9)</span>   regressed <b>1.0%</b> <span style="color:#94a3b8">(3)</span></td></tr><tr><td class="tr-name">regressed</td><td class="tr-total">27</td><td class="tr-sum">100%</td><td style="text-align:left">improved <b>29.6%</b> <span style="color:#94a3b8">(8)</span>   stuck <b>25.9%</b> <span style="color:#94a3b8">(7)</span>   churn <b>18.5%</b> <span style="color:#94a3b8">(5)</span>   breakthrough <b>14.8%</b> <span style="color:#94a3b8">(4)</span>   regressed <b>3.7%</b> <span style="color:#94a3b8">(1)</span>   end_limit <b>3.7%</b> <span style="color:#94a3b8">(1)</span>   broke <b>3.7%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">broke</td><td class="tr-total">25</td><td class="tr-sum">100%</td><td style="text-align:left">stuck <b>44.0%</b> <span style="color:#94a3b8">(11)</span>   improved <b>28.0%</b> <span style="color:#94a3b8">(7)</span>   breakthrough <b>24.0%</b> <span style="color:#94a3b8">(6)</span>   end_limit <b>4.0%</b> <span style="color:#94a3b8">(1)</span></td></tr></tbody>
	</table>
	</div>

	<div class="table-section">
	<h2>Gap label distribution — for each label, which src→tgt edges carry it</h2>
	<table class="verify">
	<thead><tr><th class="tr-name" style="text-align:left">gap label</th><th>total</th>
	<th style="text-align:left">src→tgt breakdown</th></tr></thead>
	<tbody><tr><td class="tr-name">cosmetic_refactor</td><td class="tr-total">108</td><td style="text-align:left">stay→stay <b>73.1%</b> <span style="color:#94a3b8">(79)</span>   breakthrough→stay <b>13.0%</b> <span style="color:#94a3b8">(14)</span>   improved→stuck <b>6.5%</b> <span style="color:#94a3b8">(7)</span>   stuck→stuck <b>4.6%</b> <span style="color:#94a3b8">(5)</span>   regressed→stuck <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   broke→breakthrough <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   stay→regressed <b>0.9%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">debug_failure</td><td class="tr-total">32</td><td style="text-align:left">stuck→stuck <b>28.1%</b> <span style="color:#94a3b8">(9)</span>   improved→end_limit <b>15.6%</b> <span style="color:#94a3b8">(5)</span>   stay→stay <b>12.5%</b> <span style="color:#94a3b8">(4)</span>   broke→stuck <b>12.5%</b> <span style="color:#94a3b8">(4)</span>   improved→stuck <b>6.2%</b> <span style="color:#94a3b8">(2)</span>   stay→end_limit <b>6.2%</b> <span style="color:#94a3b8">(2)</span>   churn→churn <b>6.2%</b> <span style="color:#94a3b8">(2)</span>   stuck→broke <b>3.1%</b> <span style="color:#94a3b8">(1)</span>   regressed→stuck <b>3.1%</b> <span style="color:#94a3b8">(1)</span>   stuck→end_limit <b>3.1%</b> <span style="color:#94a3b8">(1)</span>   improved→broke <b>3.1%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">edge_case</td><td class="tr-total">9</td><td style="text-align:left">stay→stay <b>66.7%</b> <span style="color:#94a3b8">(6)</span>   breakthrough→stay <b>22.2%</b> <span style="color:#94a3b8">(2)</span>   regressed→churn <b>11.1%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">execute_plan</td><td class="tr-total">15</td><td style="text-align:left">stuck→improved <b>40.0%</b> <span style="color:#94a3b8">(6)</span>   stuck→breakthrough <b>26.7%</b> <span style="color:#94a3b8">(4)</span>   stay→stay <b>6.7%</b> <span style="color:#94a3b8">(1)</span>   breakthrough→stay <b>6.7%</b> <span style="color:#94a3b8">(1)</span>   broke→improved <b>6.7%</b> <span style="color:#94a3b8">(1)</span>   broke→stuck <b>6.7%</b> <span style="color:#94a3b8">(1)</span>   broke→breakthrough <b>6.7%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">explore</td><td class="tr-total">71</td><td style="text-align:left">stuck→stuck <b>23.9%</b> <span style="color:#94a3b8">(17)</span>   start→stuck <b>23.9%</b> <span style="color:#94a3b8">(17)</span>   stay→broke <b>9.9%</b> <span style="color:#94a3b8">(7)</span>   broke→stuck <b>7.0%</b> <span style="color:#94a3b8">(5)</span>   stay→stay <b>5.6%</b> <span style="color:#94a3b8">(4)</span>   improved→end_limit <b>5.6%</b> <span style="color:#94a3b8">(4)</span>   stuck→broke <b>4.2%</b> <span style="color:#94a3b8">(3)</span>   stuck→end_limit <b>4.2%</b> <span style="color:#94a3b8">(3)</span>   improved→broke <b>4.2%</b> <span style="color:#94a3b8">(3)</span>   improved→stuck <b>2.8%</b> <span style="color:#94a3b8">(2)</span>   churn→end_limit <b>2.8%</b> <span style="color:#94a3b8">(2)</span>   start→improved <b>1.4%</b> <span style="color:#94a3b8">(1)</span>   stuck→improved <b>1.4%</b> <span style="color:#94a3b8">(1)</span>   breakthrough→broke <b>1.4%</b> <span style="color:#94a3b8">(1)</span>   broke→end_limit <b>1.4%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">fix_bug</td><td class="tr-total">107</td><td style="text-align:left">stay→stay <b>22.4%</b> <span style="color:#94a3b8">(24)</span>   stuck→stuck <b>10.3%</b> <span style="color:#94a3b8">(11)</span>   breakthrough→stay <b>9.3%</b> <span style="color:#94a3b8">(10)</span>   stuck→improved <b>9.3%</b> <span style="color:#94a3b8">(10)</span>   improved→breakthrough <b>8.4%</b> <span style="color:#94a3b8">(9)</span>   stuck→breakthrough <b>7.5%</b> <span style="color:#94a3b8">(8)</span>   improved→regressed <b>4.7%</b> <span style="color:#94a3b8">(5)</span>   improved→improved <b>3.7%</b> <span style="color:#94a3b8">(4)</span>   broke→improved <b>2.8%</b> <span style="color:#94a3b8">(3)</span>   regressed→improved <b>2.8%</b> <span style="color:#94a3b8">(3)</span>   improved→stuck <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   regressed→stuck <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   regressed→breakthrough <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   regressed→churn <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   churn→churn <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   churn→stuck <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   stuck→churn <b>1.9%</b> <span style="color:#94a3b8">(2)</span>   improved→churn <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   breakthrough→regressed <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   broke→breakthrough <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   churn→breakthrough <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   stuck→regressed <b>0.9%</b> <span style="color:#94a3b8">(1)</span>   churn→improved <b>0.9%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">implement</td><td class="tr-total">455</td><td style="text-align:left">start→breakthrough <b>68.8%</b> <span style="color:#94a3b8">(313)</span>   stay→stay <b>7.7%</b> <span style="color:#94a3b8">(35)</span>   start→improved <b>7.3%</b> <span style="color:#94a3b8">(33)</span>   start→stuck <b>3.7%</b> <span style="color:#94a3b8">(17)</span>   breakthrough→stay <b>3.3%</b> <span style="color:#94a3b8">(15)</span>   stuck→stuck <b>1.8%</b> <span style="color:#94a3b8">(8)</span>   stuck→improved <b>1.3%</b> <span style="color:#94a3b8">(6)</span>   improved→stuck <b>1.1%</b> <span style="color:#94a3b8">(5)</span>   breakthrough→regressed <b>0.7%</b> <span style="color:#94a3b8">(3)</span>   stuck→breakthrough <b>0.4%</b> <span style="color:#94a3b8">(2)</span>   improved→regressed <b>0.4%</b> <span style="color:#94a3b8">(2)</span>   regressed→improved <b>0.4%</b> <span style="color:#94a3b8">(2)</span>   stuck→regressed <b>0.4%</b> <span style="color:#94a3b8">(2)</span>   churn→stuck <b>0.4%</b> <span style="color:#94a3b8">(2)</span>   broke→improved <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   stuck→broke <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   regressed→stuck <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   broke→breakthrough <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   regressed→breakthrough <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   improved→breakthrough <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   breakthrough→broke <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   churn→churn <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   improved→broke <b>0.2%</b> <span style="color:#94a3b8">(1)</span>   stuck→churn <b>0.2%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">infra_confusion</td><td class="tr-total">9</td><td style="text-align:left">stay→stay <b>44.4%</b> <span style="color:#94a3b8">(4)</span>   stay→end_submit <b>22.2%</b> <span style="color:#94a3b8">(2)</span>   stuck→breakthrough <b>11.1%</b> <span style="color:#94a3b8">(1)</span>   stay→broke <b>11.1%</b> <span style="color:#94a3b8">(1)</span>   improved→breakthrough <b>11.1%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">optimize_perf</td><td class="tr-total">87</td><td style="text-align:left">stay→stay <b>48.3%</b> <span style="color:#94a3b8">(42)</span>   breakthrough→stay <b>21.8%</b> <span style="color:#94a3b8">(19)</span>   stuck→breakthrough <b>4.6%</b> <span style="color:#94a3b8">(4)</span>   improved→stuck <b>3.4%</b> <span style="color:#94a3b8">(3)</span>   stuck→stuck <b>3.4%</b> <span style="color:#94a3b8">(3)</span>   improved→regressed <b>3.4%</b> <span style="color:#94a3b8">(3)</span>   improved→improved <b>2.3%</b> <span style="color:#94a3b8">(2)</span>   improved→breakthrough <b>2.3%</b> <span style="color:#94a3b8">(2)</span>   stay→regressed <b>2.3%</b> <span style="color:#94a3b8">(2)</span>   broke→stuck <b>1.1%</b> <span style="color:#94a3b8">(1)</span>   breakthrough→regressed <b>1.1%</b> <span style="color:#94a3b8">(1)</span>   regressed→stuck <b>1.1%</b> <span style="color:#94a3b8">(1)</span>   regressed→breakthrough <b>1.1%</b> <span style="color:#94a3b8">(1)</span>   regressed→churn <b>1.1%</b> <span style="color:#94a3b8">(1)</span>   regressed→improved <b>1.1%</b> <span style="color:#94a3b8">(1)</span>   regressed→broke <b>1.1%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">revert</td><td class="tr-total">25</td><td style="text-align:left">stay→stay <b>60.0%</b> <span style="color:#94a3b8">(15)</span>   stuck→stuck <b>32.0%</b> <span style="color:#94a3b8">(8)</span>   improved→stuck <b>4.0%</b> <span style="color:#94a3b8">(1)</span>   churn→breakthrough <b>4.0%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">rewrite</td><td class="tr-total">54</td><td style="text-align:left">stuck→stuck <b>14.8%</b> <span style="color:#94a3b8">(8)</span>   churn→churn <b>11.1%</b> <span style="color:#94a3b8">(6)</span>   stuck→improved <b>9.3%</b> <span style="color:#94a3b8">(5)</span>   breakthrough→stay <b>7.4%</b> <span style="color:#94a3b8">(4)</span>   improved→churn <b>7.4%</b> <span style="color:#94a3b8">(4)</span>   improved→regressed <b>7.4%</b> <span style="color:#94a3b8">(4)</span>   improved→breakthrough <b>5.6%</b> <span style="color:#94a3b8">(3)</span>   broke→improved <b>3.7%</b> <span style="color:#94a3b8">(2)</span>   improved→stuck <b>3.7%</b> <span style="color:#94a3b8">(2)</span>   churn→broke <b>3.7%</b> <span style="color:#94a3b8">(2)</span>   stuck→breakthrough <b>3.7%</b> <span style="color:#94a3b8">(2)</span>   broke→breakthrough <b>3.7%</b> <span style="color:#94a3b8">(2)</span>   regressed→improved <b>3.7%</b> <span style="color:#94a3b8">(2)</span>   stuck→broke <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   breakthrough→regressed <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   regressed→stuck <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   stay→broke <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   regressed→churn <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   regressed→regressed <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   churn→stuck <b>1.9%</b> <span style="color:#94a3b8">(1)</span>   stuck→churn <b>1.9%</b> <span style="color:#94a3b8">(1)</span></td></tr><tr><td class="tr-name">verify</td><td class="tr-total">387</td><td style="text-align:left">breakthrough→end_submit <b>73.9%</b> <span style="color:#94a3b8">(286)</span>   stay→end_submit <b>10.9%</b> <span style="color:#94a3b8">(42)</span>   stay→stay <b>5.2%</b> <span style="color:#94a3b8">(20)</span>   stay→end_limit <b>2.6%</b> <span style="color:#94a3b8">(10)</span>   stuck→end_limit <b>2.3%</b> <span style="color:#94a3b8">(9)</span>   stuck→end_submit <b>1.6%</b> <span style="color:#94a3b8">(6)</span>   improved→end_limit <b>1.0%</b> <span style="color:#94a3b8">(4)</span>   breakthrough→stay <b>0.8%</b> <span style="color:#94a3b8">(3)</span>   stuck→stuck <b>0.8%</b> <span style="color:#94a3b8">(3)</span>   churn→end_limit <b>0.5%</b> <span style="color:#94a3b8">(2)</span>   regressed→end_limit <b>0.3%</b> <span style="color:#94a3b8">(1)</span>   improved→end_submit <b>0.3%</b> <span style="color:#94a3b8">(1)</span></td></tr></tbody>
	</table>
	</div>

	<script>
	const M = {"canvas": {"w": 760, "h": 760}, "nodes": [{"id": "start", "label": "▶", "state_type": "start", "color": "#9E9E9E", "definition": "Virtual starting state — before any solution.py was written.", "visits": 381, "out_total": 381, "nlp_out": {"m1_correctness": 48.6, "m1_efficiency": 29.7, "m2": 48.6, "m3_existing": 54.1, "m4": 16.3, "m7": 2.6, "m3_sample": 12.9, "m8_inline": 5.0, "m9": 9.4, "m6": 16.3, "m3_own": 0.3, "m1_clarity": 0.5, "m8_lang": 1.3}, "nlp_out_total": 381, "nlp_in": {}, "nlp_in_total": 0, "gap_out": {"implement": 95.3, "explore": 4.7}, "gap_out_total": 381, "gap_in": {}, "gap_in_total": 0, "x": 80.0, "y": 380.0}, {"id": "improved", "label": "improved", "state_type": "improved", "color": "#2196F3", "definition": "Gained tests, lost none (but not yet all-pass).", "visits": 84, "out_total": 84, "nlp_out": {"m3_existing": 36.9, "m1_efficiency": 47.6, "m2": 72.6, "m4": 34.5, "m8_inline": 9.5, "m9": 28.6, "m7": 3.6, "m1_correctness": 75.0, "m3_sample": 38.1, "m6": 21.4, "m8_lang": 8.3, "m3_own": 8.3}, "nlp_out_total": 84, "nlp_in": {"m3_existing": 28.6, "m9": 17.9, "m1_efficiency": 40.5, "m2": 83.3, "m4": 38.1, "m1_correctness": 79.8, "m3_sample": 42.9, "m8_inline": 6.0, "m6": 16.7, "m3_own": 7.1, "m7": 1.2, "m8_lang": 2.4}, "nlp_in_total": 84, "gap_out": {"fix_bug": 25.0, "rewrite": 15.5, "optimize_perf": 11.9, "implement": 10.7, "explore": 10.7, "debug_failure": 9.5, "cosmetic_refactor": 8.3, "verify": 6.0, "revert": 1.2, "infra_confusion": 1.2}, "gap_out_total": 84, "gap_in": {"implement": 50.0, "fix_bug": 25.0, "rewrite": 10.7, "execute_plan": 8.3, "optimize_perf": 3.6, "explore": 2.4}, "gap_in_total": 84, "x": 260.0, "y": 170.0}, {"id": "churn", "label": "churn", "state_type": "churn", "color": "#9C27B0", "definition": "Gained some tests AND lost others — mixed regression+improvement.", "visits": 25, "out_total": 25, "nlp_out": {"m1_correctness": 84.0, "m2": 92.0, "m3_sample": 12.0, "m4": 16.0, "m9": 20.0, "m1_efficiency": 64.0, "m6": 44.0, "m7": 12.0, "m8_inline": 12.0, "m3_existing": 40.0, "m8_lang": 4.0}, "nlp_out_total": 25, "nlp_in": {"m1_correctness": 84.0, "m1_efficiency": 76.0, "m2": 100.0, "m3_sample": 16.0, "m4": 20.0, "m3_existing": 28.0, "m6": 40.0, "m9": 12.0, "m7": 4.0, "m8_inline": 8.0, "m8_lang": 8.0}, "nlp_in_total": 25, "gap_out": {"rewrite": 36.0, "fix_bug": 24.0, "implement": 12.0, "debug_failure": 8.0, "explore": 8.0, "verify": 8.0, "revert": 4.0}, "gap_out_total": 25, "gap_in": {"rewrite": 48.0, "fix_bug": 28.0, "debug_failure": 8.0, "implement": 8.0, "edge_case": 4.0, "optimize_perf": 4.0}, "gap_in_total": 25, "x": 365.0, "y": 170.0}, {"id": "stuck", "label": "stuck", "state_type": "stuck", "color": "#FF9800", "definition": "Identical test tuple — no change from the previous snapshot.", "visits": 153, "out_total": 153, "nlp_out": {"m1_correctness": 68.6, "m2": 77.8, "m4": 37.9, "m3_existing": 13.7, "m9": 17.0, "m8_inline": 3.9, "m1_efficiency": 46.4, "m3_sample": 45.1, "m3_own": 3.3, "m8_lang": 6.5, "m6": 18.3, "m7": 2.6}, "nlp_out_total": 153, "nlp_in": {"m1_correctness": 74.5, "m1_efficiency": 52.9, "m2": 85.0, "m3_existing": 14.4, "m4": 37.9, "m9": 22.2, "m3_sample": 39.2, "m8_inline": 2.6, "m8_lang": 3.3, "m6": 17.0, "m3_own": 3.3, "m7": 2.0}, "nlp_in_total": 153, "gap_out": {"fix_bug": 20.9, "explore": 15.7, "implement": 13.1, "verify": 11.8, "rewrite": 11.1, "debug_failure": 7.2, "execute_plan": 6.5, "revert": 5.2, "optimize_perf": 4.6, "cosmetic_refactor": 3.3, "infra_confusion": 0.7}, "gap_out_total": 153, "gap_in": {"explore": 26.8, "implement": 21.6, "fix_bug": 11.1, "debug_failure": 10.5, "cosmetic_refactor": 8.5, "rewrite": 7.8, "revert": 5.9, "optimize_perf": 5.2, "verify": 2.0, "execute_plan": 0.7}, "gap_in_total": 153, "x": 260.0, "y": 380.0}, {"id": "breakthrough", "label": "breakthrough", "state_type": "breakthrough", "color": "#00BCD4", "definition": "Some failing → all passing.", "visits": 362, "out_total": 362, "nlp_out": {"m1_correctness": 46.1, "m1_efficiency": 19.3, "m3_existing": 45.3, "m3_sample": 16.9, "m8_inline": 1.4, "m2": 26.5, "m6": 18.0, "m8_lang": 0.8, "m7": 3.3, "m3_own": 0.8, "m9": 5.5, "m4": 5.8}, "nlp_out_total": 362, "nlp_in": {"m1_correctness": 43.4, "m1_efficiency": 27.1, "m2": 43.9, "m3_existing": 55.0, "m4": 15.7, "m7": 4.1, "m3_sample": 16.9, "m8_inline": 6.1, "m9": 6.1, "m6": 18.5, "m3_own": 0.6, "m1_clarity": 0.6, "m8_lang": 1.1}, "nlp_in_total": 362, "gap_out": {"verify": 79.8, "optimize_perf": 5.5, "implement": 5.2, "cosmetic_refactor": 3.9, "fix_bug": 3.0, "rewrite": 1.4, "edge_case": 0.6, "execute_plan": 0.3, "explore": 0.3}, "gap_out_total": 362, "gap_in": {"implement": 87.8, "fix_bug": 5.8, "optimize_perf": 1.9, "rewrite": 1.9, "execute_plan": 1.4, "infra_confusion": 0.6, "cosmetic_refactor": 0.3, "revert": 0.3}, "gap_in_total": 362, "x": 260.0, "y": 590.0}, {"id": "stay", "label": "stay", "state_type": "stay", "color": "#4CAF50", "definition": "All-pass → all-pass (model rewrote a working solution).", "visits": 302, "out_total": 302, "nlp_out": {"m1_efficiency": 72.5, "m3_existing": 28.1, "m4": 16.9, "m1_correctness": 81.8, "m2": 81.1, "m9": 7.6, "m8_inline": 1.7, "m6": 65.9, "m7": 9.3, "m3_sample": 55.3, "m3_own": 6.6, "m8_lang": 5.3}, "nlp_out_total": 302, "nlp_in": {"m1_efficiency": 76.2, "m3_existing": 30.5, "m3_sample": 66.9, "m4": 20.2, "m8_inline": 2.0, "m9": 8.6, "m1_correctness": 85.1, "m2": 86.1, "m6": 71.9, "m7": 10.6, "m3_own": 6.6, "m8_lang": 4.6}, "nlp_in_total": 302, "gap_out": {"cosmetic_refactor": 26.5, "verify": 23.8, "optimize_perf": 14.6, "implement": 11.6, "fix_bug": 7.9, "revert": 5.0, "explore": 3.6, "infra_confusion": 2.3, "edge_case": 2.0, "debug_failure": 2.0, "execute_plan": 0.3, "rewrite": 0.3}, "gap_out_total": 302, "gap_in": {"cosmetic_refactor": 30.8, "optimize_perf": 20.2, "implement": 16.6, "fix_bug": 11.3, "verify": 7.6, "revert": 5.0, "edge_case": 2.6, "explore": 1.3, "debug_failure": 1.3, "infra_confusion": 1.3, "rewrite": 1.3, "execute_plan": 0.7}, "gap_in_total": 302, "x": 470.0, "y": 590.0}, {"id": "regressed", "label": "regressed", "state_type": "regressed", "color": "#F44336", "definition": "Lost tests, gained none (but not all-fail).", "visits": 27, "out_total": 27, "nlp_out": {"m1_efficiency": 55.6, "m2": 92.6, "m3_sample": 51.9, "m3_own": 14.8, "m4": 14.8, "m9": 25.9, "m1_correctness": 66.7, "m3_existing": 11.1, "m6": 40.7, "m7": 7.4, "m8_inline": 3.7}, "nlp_out_total": 27, "nlp_in": {"m1_correctness": 88.9, "m1_efficiency": 63.0, "m3_existing": 33.3, "m3_sample": 55.6, "m3_own": 14.8, "m4": 37.0, "m8_lang": 3.7, "m8_inline": 3.7, "m9": 40.7, "m2": 96.3, "m6": 40.7, "m7": 7.4}, "nlp_in_total": 27, "gap_out": {"fix_bug": 33.3, "rewrite": 18.5, "optimize_perf": 18.5, "implement": 14.8, "cosmetic_refactor": 3.7, "debug_failure": 3.7, "edge_case": 3.7, "verify": 3.7}, "gap_out_total": 27, "gap_in": {"implement": 25.9, "fix_bug": 25.9, "optimize_perf": 22.2, "rewrite": 22.2, "cosmetic_refactor": 3.7}, "gap_in_total": 27, "x": 470.0, "y": 170.0}, {"id": "broke", "label": "broke", "state_type": "broke", "color": "#7B0000", "definition": "Had some passing → zero passing (or compile/runtime failure).", "visits": 25, "out_total": 25, "nlp_out": {"m3_existing": 12.0, "m4": 52.0, "m8_lang": 4.0, "m9": 8.0, "m1_correctness": 56.0, "m2": 64.0, "m3_sample": 44.0, "m6": 12.0, "m1_efficiency": 32.0}, "nlp_out_total": 25, "nlp_in": {"m1_correctness": 80.0, "m3_existing": 12.0, "m4": 64.0, "m1_efficiency": 68.0, "m2": 88.0, "m6": 32.0, "m3_sample": 20.0, "m9": 16.0, "m3_own": 4.0, "m8_lang": 8.0, "m7": 4.0}, "nlp_in_total": 25, "gap_out": {"explore": 24.0, "rewrite": 16.0, "fix_bug": 16.0, "debug_failure": 16.0, "execute_plan": 12.0, "implement": 8.0, "optimize_perf": 4.0, "cosmetic_refactor": 4.0}, "gap_out_total": 25, "gap_in": {"explore": 56.0, "rewrite": 16.0, "implement": 12.0, "debug_failure": 8.0, "infra_confusion": 4.0, "optimize_perf": 4.0}, "gap_in_total": 25, "x": 470.0, "y": 380.0}, {"id": "end_submit", "label": "SUBMIT", "state_type": "end_submit", "color": "#607D8B", "definition": "Model submitted itself (ran COMPLETE_TASK_AND_SUBMIT_FINAL_OUTPUT).", "visits": 339, "out_total": 0, "nlp_out": {}, "nlp_out_total": 0, "nlp_in": {"m1_correctness": 43.1, "m1_efficiency": 13.9, "m3_existing": 44.2, "m3_sample": 6.5, "m8_inline": 0.9, "m2": 19.8, "m6": 10.9, "m8_lang": 0.9, "m7": 1.8, "m3_own": 0.6, "m9": 4.1}, "nlp_in_total": 339, "gap_out": {}, "gap_out_total": 0, "gap_in": {"verify": 99.4, "infra_confusion": 0.6}, "gap_in_total": 337, "x": 680.0, "y": 290.0}, {"id": "end_limit", "label": "LIMIT", "state_type": "end_limit", "color": "#455A64", "definition": "Model force-stopped at the turn limit / wall-clock timeout.", "visits": 42, "out_total": 0, "nlp_out": {}, "nlp_out_total": 0, "nlp_in": {"m4": 7.1, "m8_lang": 23.8, "m3_existing": 40.5, "m8_inline": 9.5, "m9": 33.3, "m1_correctness": 33.3, "m1_efficiency": 21.4, "m2": 26.2, "m6": 16.7, "m3_sample": 2.4, "m7": 2.4}, "nlp_in_total": 42, "gap_out": {}, "gap_out_total": 0, "gap_in": {"verify": 59.1, "explore": 22.7, "debug_failure": 18.2}, "gap_in_total": 44, "x": 680.0, "y": 470.0}], "edges": [{"id": "m0", "source": "start", "target": "breakthrough", "color": "#9E9E9E", "count": 313, "out_pct": 82.2, "label": "313", "n_unique_chains": 15, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 244, "pct": 78.0, "examples": [{"problem": "claude-sonnet-4-6/1883/1883_B", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc303/abc303_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc305/abc305_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc305/abc305_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc308/abc308_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc312/abc312_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc313/abc313_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc315/abc315_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc318/abc318_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc322/abc322_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc324/abc324_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc326/abc326_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc329/abc329_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc330/abc330_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc334/abc334_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc335/abc335_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc341/abc341_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc342/abc342_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc351/abc351_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc354/abc354_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc355/abc355_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc356/abc356_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc357/abc357_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc358/abc358_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc360/abc360_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc366/abc366_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc373/abc373_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc383/abc383_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc384/abc384_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc384/abc384_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc387/abc387_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc387/abc387_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc389/abc389_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc391/abc391_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc391/abc391_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc391/abc391_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc392/abc392_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc395/abc395_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc396/abc396_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc397/abc397_b", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 31, "pct": 9.9, "examples": [{"problem": "claude-sonnet-4-6/abc302/abc302_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc308/abc308_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc310/abc310_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc311/abc311_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc319/abc319_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc324/abc324_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc333/abc333_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc334/abc334_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc340/abc340_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc341/abc341_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc344/abc344_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc345/abc345_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc357/abc357_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc359/abc359_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc365/abc365_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc381/abc381_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc381/abc381_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc385/abc385_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc390/abc390_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc392/abc392_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc395/abc395_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/arc191/arc191_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/arc194/arc194_a", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc303/abc303_a", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc334/abc334_c", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc357/abc357_d", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc360/abc360_a", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc302/abc302_c", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc333/abc333_b", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc385/abc385_e", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc398/abc398_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "write"], "label": "inspect test file → inspect test file → write solution.py", "count": 14, "pct": 4.5, "examples": [{"problem": "claude-sonnet-4-6/abc310/abc310_f", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc325/abc325_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc332/abc332_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc358/abc358_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc371/abc371_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc388/abc388_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc388/abc388_f", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc400/abc400_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc311/abc311_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc312/abc312_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc333/abc333_c", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc340/abc340_c", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc384/abc384_c", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_e", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inline_script", "write"], "label": "inline_script → write solution.py", "count": 7, "pct": 2.2, "examples": [{"problem": "gpt-5.5/abc301/abc301_f", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc359/abc359_c", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc397/abc397_e", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/abc400/abc400_d", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/arc194/arc194_b", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/arc194/arc194_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc356/abc356_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → write solution.py", "count": 5, "pct": 1.6, "examples": [{"problem": "claude-sonnet-4-6/abc312/abc312_d", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc333/abc333_c", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc393/abc393_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/arc194/arc194_e", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc395/abc395_c", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 2, "pct": 0.6, "examples": [{"problem": "gpt-5.5/arc184/arc184_d", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/arc194/arc194_e", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inline_script", "inline_script", "write"], "label": "inline_script → inline_script → write solution.py", "count": 2, "pct": 0.6, "examples": [{"problem": "gpt-5.5/arc186/arc186_b", "snap_from": -1, "snap_to": 0}, {"problem": "gpt-5.5/arc196/arc196_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inspect test file → inspect test file → inspect test file → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "claude-sonnet-4-6/abc359/abc359_c", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inline_script", "inline_script", "inline_script", "write"], "label": "inspect test file → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "claude-sonnet-4-6/arc186/arc186_b", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inline_script", "inline_script", "inline_script", "write"], "label": "inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "gpt-5.5/arc193/arc193_a", "snap_from": -1, "snap_to": 0}]}, {"chain": ["write", "write"], "label": "write solution.py → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc308/abc308_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["write_script", "write_script", "write"], "label": "write helper script → write helper script → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc397/abc397_e", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "inspect", "wc", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → inspect test file → wc (output size) → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc191/arc191_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["write_script", "write"], "label": "write helper script → write solution.py", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc193/arc193_a", "snap_from": -1, "snap_to": 0}]}], "label_dist": [{"label": "implement", "count": 313, "pct": 100.0}], "n_labeled": 313, "nlp_rates": {"m1_correctness": 41, "m1_efficiency": 25, "m2": 40, "m3_existing": 58, "m4": 11, "m7": 3, "m3_sample": 12, "m8_inline": 5, "m9": 4, "m6": 18, "m3_own": 0, "m1_clarity": 1, "m8_lang": 1}}, {"id": "m1", "source": "breakthrough", "target": "end_submit", "color": "#00BCD4", "count": 286, "out_pct": 79.0, "label": "286", "n_unique_chains": 12, "breakdown": [{"chain": ["submit"], "label": "submit", "count": 154, "pct": 53.8, "examples": [{"problem": "claude-sonnet-4-6/1883/1883_B", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc302/abc302_a", "snap_from": 1, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc302/abc302_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc303/abc303_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc305/abc305_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc305/abc305_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc310/abc310_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc312/abc312_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc312/abc312_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc313/abc313_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc315/abc315_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc318/abc318_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc322/abc322_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc324/abc324_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc325/abc325_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc326/abc326_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc329/abc329_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc330/abc330_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc332/abc332_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc333/abc333_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc334/abc334_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc335/abc335_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc340/abc340_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc341/abc341_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc341/abc341_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc342/abc342_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc344/abc344_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc351/abc351_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc354/abc354_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc355/abc355_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc357/abc357_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc357/abc357_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc358/abc358_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc359/abc359_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc360/abc360_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc371/abc371_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc373/abc373_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc381/abc381_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc383/abc383_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc384/abc384_a", "snap_from": 0, "snap_to": -1}]}, {"chain": ["test", "submit"], "label": "run ./test.sh → submit", "count": 114, "pct": 39.9, "examples": [{"problem": "claude-sonnet-4-6/abc308/abc308_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc308/abc308_e", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc310/abc310_f", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc311/abc311_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc319/abc319_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc333/abc333_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc334/abc334_c", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc345/abc345_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc356/abc356_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc358/abc358_e", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc359/abc359_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc365/abc365_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc366/abc366_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc381/abc381_e", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc385/abc385_e", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc388/abc388_f", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc391/abc391_e", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc398/abc398_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/arc186/arc186_b", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/arc194/arc194_a", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/arc194/arc194_b", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc302/abc302_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc305/abc305_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc313/abc313_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc315/abc315_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc318/abc318_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc322/abc322_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc326/abc326_b", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc329/abc329_b", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc330/abc330_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc333/abc333_b", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc335/abc335_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc336/abc336_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc341/abc341_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc341/abc341_d", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc351/abc351_a", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc356/abc356_d", "snap_from": 0, "snap_to": -1}]}, {"chain": ["run_sample", "test", "submit"], "label": "run solution.py (sample) → run ./test.sh → submit", "count": 5, "pct": 1.7, "examples": [{"problem": "claude-sonnet-4-6/abc324/abc324_d", "snap_from": 0, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc397/abc397_e", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/1883/1883_B", "snap_from": 0, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc354/abc354_d", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/arc194/arc194_e", "snap_from": 0, "snap_to": -1}]}, {"chain": ["test", "test", "submit"], "label": "run ./test.sh → run ./test.sh → submit", "count": 3, "pct": 1.0, "examples": [{"problem": "claude-sonnet-4-6/abc400/abc400_d", "snap_from": 0, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc308/abc308_e", "snap_from": 0, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc381/abc381_e", "snap_from": 0, "snap_to": -1}]}, {"chain": ["inline_script", "test", "submit"], "label": "inline_script → run ./test.sh → submit", "count": 2, "pct": 0.7, "examples": [{"problem": "gpt-5.5/abc389/abc389_g", "snap_from": 3, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc345/abc345_d", "snap_from": 4, "snap_to": -1}]}, {"chain": ["run_sample", "run_sample", "test", "submit"], "label": "run solution.py (sample) → run solution.py (sample) → run ./test.sh → submit", "count": 2, "pct": 0.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc384/abc384_c", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc383/abc383_c", "snap_from": 1, "snap_to": -1}]}, {"chain": ["run_sample", "test", "test", "submit"], "label": "run solution.py (sample) → run ./test.sh → run ./test.sh → submit", "count": 1, "pct": 0.3, "examples": [{"problem": "claude-sonnet-4-6/abc393/abc393_e", "snap_from": 0, "snap_to": -1}]}, {"chain": ["run_sample", "test", "test", "test", "submit"], "label": "run solution.py (sample) → run ./test.sh → run ./test.sh → run ./test.sh → submit", "count": 1, "pct": 0.3, "examples": [{"problem": "claude-sonnet-4-6/arc196/arc196_d", "snap_from": 2, "snap_to": -1}]}, {"chain": ["write_script", "test", "submit"], "label": "write helper script → run ./test.sh → submit", "count": 1, "pct": 0.3, "examples": [{"problem": "gemma_4_26b_a4b_it/abc305/abc305_b", "snap_from": 0, "snap_to": -1}]}, {"chain": ["inspect", "inspect", "inspect", "test", "submit"], "label": "inspect test file → inspect test file → inspect test file → run ./test.sh → submit", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc311/abc311_d", "snap_from": 0, "snap_to": -1}]}, {"chain": ["inspect", "test", "submit"], "label": "inspect test file → run ./test.sh → submit", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc365/abc365_d", "snap_from": 2, "snap_to": -1}]}, {"chain": ["inspect", "inspect", "inspect", "run_sample", "run_sample", "test", "submit"], "label": "inspect test file → inspect test file → inspect test file → run solution.py (sample) → run solution.py (sample) → run ./test.sh → submit", "count": 1, "pct": 0.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc393/abc393_e", "snap_from": 0, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 286, "pct": 100.0}], "n_labeled": 286, "nlp_rates": {"m1_correctness": 37, "m1_efficiency": 7, "m3_existing": 48, "m3_sample": 6, "m8_inline": 1, "m2": 12, "m6": 7, "m8_lang": 0, "m7": 1, "m3_own": 0, "m9": 4}}, {"id": "m2", "source": "start", "target": "stuck", "color": "#9E9E9E", "count": 34, "out_pct": 8.9, "label": "34", "n_unique_chains": 10, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 24, "pct": 70.6, "examples": [{"problem": "gpt-5.5/abc389/abc389_g", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc337/abc337_e", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc354/abc354_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc359/abc359_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc378/abc378_g", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc382/abc382_g", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc384/abc384_c", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_b", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_a", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc196/arc196_a", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc196/arc196_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc351/abc351_a", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc354/abc354_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc365/abc365_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc382/abc382_g", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc391/abc391_e", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc184/arc184_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": -1, "snap_to": 0}]}, {"chain": ["write", "write"], "label": "write solution.py → write solution.py", "count": 2, "pct": 5.9, "examples": [{"problem": "gemma_4_26b_a4b_it/abc389/abc389_g", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc195/arc195_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "claude-sonnet-4-6/abc301/abc301_f", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "write"], "label": "inspect test file → inspect test file → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "claude-sonnet-4-6/abc337/abc337_e", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "wc", "write"], "label": "inspect test file → wc (output size) → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "claude-sonnet-4-6/arc193/arc193_a", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "claude-sonnet-4-6/arc196/arc196_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inline_script", "inline_script", "write"], "label": "inline_script → inline_script → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "gpt-5.5/arc194/arc194_a", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc301/abc301_f", "snap_from": -1, "snap_to": 0}]}, {"chain": ["write_script", "write_script", "write"], "label": "write helper script → write helper script → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": -1, "snap_to": 0}]}, {"chain": ["write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write"], "label": "write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc194/arc194_d", "snap_from": -1, "snap_to": 0}]}], "label_dist": [{"label": "implement", "count": 17, "pct": 50.0}, {"label": "explore", "count": 17, "pct": 50.0}], "n_labeled": 34, "nlp_rates": {"m1_correctness": 76, "m1_efficiency": 44, "m2": 88, "m3_existing": 21, "m4": 44, "m9": 41, "m3_sample": 9, "m8_inline": 3, "m8_lang": 9, "m6": 6}}, {"id": "m3", "source": "stuck", "target": "stuck", "color": "#FF9800", "count": 72, "out_pct": 47.1, "label": "72", "n_unique_chains": 5, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 66, "pct": 91.7, "examples": [{"problem": "claude-sonnet-4-6/abc301/abc301_f", "snap_from": 0, "snap_to": 1}, {"problem": "claude-sonnet-4-6/abc337/abc337_e", "snap_from": 0, "snap_to": 1}, {"problem": "claude-sonnet-4-6/abc337/abc337_e", "snap_from": 1, "snap_to": 2}, {"problem": "claude-sonnet-4-6/abc389/abc389_g", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc359/abc359_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc382/abc382_g", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc382/abc382_g", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc382/abc382_g", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc382/abc382_g", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc389/abc389_g", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc395/abc395_c", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 14, "snap_to": 15}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 15, "snap_to": 16}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 16, "snap_to": 17}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 17, "snap_to": 18}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_a", "snap_from": 0, "snap_to": 1}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 3, "pct": 4.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc365/abc365_d", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 5, "snap_to": 6}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 10, "snap_to": 11}]}, {"chain": ["write_script", "write"], "label": "write helper script → write solution.py", "count": 1, "pct": 1.4, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 9, "snap_to": 10}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 1, "pct": 1.4, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 6, "snap_to": 7}]}, {"chain": ["inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 1.4, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 9, "snap_to": 10}]}], "label_dist": [{"label": "explore", "count": 17, "pct": 23.6}, {"label": "fix_bug", "count": 11, "pct": 15.3}, {"label": "debug_failure", "count": 9, "pct": 12.5}, {"label": "revert", "count": 8, "pct": 11.1}, {"label": "rewrite", "count": 8, "pct": 11.1}, {"label": "implement", "count": 8, "pct": 11.1}, {"label": "cosmetic_refactor", "count": 5, "pct": 6.9}, {"label": "verify", "count": 3, "pct": 4.2}, {"label": "optimize_perf", "count": 3, "pct": 4.2}], "n_labeled": 72, "nlp_rates": {"m1_correctness": 75, "m2": 90, "m4": 36, "m3_existing": 10, "m9": 18, "m8_inline": 1, "m1_efficiency": 57, "m3_sample": 51, "m3_own": 3, "m8_lang": 1, "m6": 18, "m7": 3}}, {"id": "m4", "source": "stuck", "target": "end_limit", "color": "#FF9800", "count": 12, "out_pct": 7.8, "label": "12", "n_unique_chains": 4, "breakdown": [{"chain": [], "label": "(no useful command)", "count": 8, "pct": 66.7, "examples": [{"problem": "claude-sonnet-4-6/abc389/abc389_g", "snap_from": 4, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc354/abc354_d", "snap_from": 0, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc378/abc378_g", "snap_from": 4, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 11, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_a", "snap_from": 6, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc301/abc301_f", "snap_from": 3, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/arc195/arc195_d", "snap_from": 2, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": 5, "snap_to": -1}]}, {"chain": ["write_script"], "label": "write helper script", "count": 2, "pct": 16.7, "examples": [{"problem": "claude-sonnet-4-6/abc301/abc301_f", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 9, "snap_to": -1}]}, {"chain": ["inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script"], "label": "inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script", "count": 1, "pct": 8.3, "examples": [{"problem": "claude-sonnet-4-6/abc337/abc337_e", "snap_from": 2, "snap_to": -1}]}, {"chain": ["write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script"], "label": "write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script", "count": 1, "pct": 8.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc382/abc382_g", "snap_from": 1, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 9, "pct": 69.2}, {"label": "explore", "count": 3, "pct": 23.1}, {"label": "debug_failure", "count": 1, "pct": 7.7}], "n_labeled": 13, "nlp_rates": {"m4": 8, "m8_lang": 25, "m3_existing": 8, "m8_inline": 8, "m9": 25, "m1_correctness": 17, "m1_efficiency": 8, "m2": 17, "m6": 8}}, {"id": "m5", "source": "start", "target": "improved", "color": "#9E9E9E", "count": 34, "out_pct": 8.9, "label": "34", "n_unique_chains": 6, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 20, "pct": 58.8, "examples": [{"problem": "claude-sonnet-4-6/abc302/abc302_a", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/arc196/arc196_a", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc395/abc395_c", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc397/abc397_e", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc192/arc192_b", "snap_from": -1, "snap_to": 0}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_e", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc310/abc310_f", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc313/abc313_a", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc390/abc390_e", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc398/abc398_g", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc186/arc186_b", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc186/arc186_e", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc192/arc192_b", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "write"], "label": "inspect test file → inspect test file → write solution.py", "count": 7, "pct": 20.6, "examples": [{"problem": "claude-sonnet-4-6/abc389/abc389_g", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc397/abc397_e", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc398/abc398_g", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/abc400/abc400_g", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 4, "pct": 11.8, "examples": [{"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/arc192/arc192_b", "snap_from": -1, "snap_to": 0}, {"problem": "claude-sonnet-4-6/arc194/arc194_b", "snap_from": -1, "snap_to": 0}, {"problem": "qwen3p6_35b_a3b_fp8/abc383/abc383_c", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "gpt-5.5/arc192/arc192_b", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "test", "inspect", "inspect", "inspect", "wc", "write"], "label": "inspect test file → inspect test file → run ./test.sh → inspect test file → inspect test file → inspect test file → wc (output size) → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc337/abc337_e", "snap_from": -1, "snap_to": 0}]}, {"chain": ["inspect", "inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 2.9, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc345/abc345_d", "snap_from": -1, "snap_to": 0}]}], "label_dist": [{"label": "implement", "count": 33, "pct": 97.1}, {"label": "explore", "count": 1, "pct": 2.9}], "n_labeled": 34, "nlp_rates": {"m3_existing": 47, "m9": 24, "m1_efficiency": 56, "m2": 85, "m4": 35, "m1_correctness": 88, "m3_sample": 21, "m8_inline": 9, "m6": 15}}, {"id": "m6", "source": "improved", "target": "breakthrough", "color": "#2196F3", "count": 16, "out_pct": 19.0, "label": "16", "n_unique_chains": 9, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 8, "pct": 50.0, "examples": [{"problem": "claude-sonnet-4-6/abc302/abc302_a", "snap_from": 0, "snap_to": 1}, {"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": 1, "snap_to": 2}, {"problem": "claude-sonnet-4-6/abc397/abc397_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc397/abc397_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_e", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc310/abc310_f", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "claude-sonnet-4-6/arc194/arc194_b", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_sample", "run_sample", "test", "test", "inspect", "inspect", "inline_script", "inline_script", "inline_script", "write"], "label": "run solution.py (sample) → run solution.py (sample) → run ./test.sh → run ./test.sh → inspect test file → inspect test file → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "claude-sonnet-4-6/arc196/arc196_d", "snap_from": 1, "snap_to": 2}]}, {"chain": ["inspect", "inline_script", "inline_script", "inline_script", "inspect", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inspect test file → inline_script → inline_script → inline_script → inspect test file → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "gpt-5.5/arc192/arc192_b", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_sample", "write_script", "write"], "label": "run solution.py (sample) → write helper script → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "gemma_4_26b_a4b_it/arc194/arc194_d", "snap_from": 2, "snap_to": 3}]}, {"chain": ["test", "inspect", "write"], "label": "run ./test.sh → inspect test file → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc313/abc313_a", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_sample", "write"], "label": "run solution.py (sample) → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc383/abc383_c", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc390/abc390_e", "snap_from": 4, "snap_to": 5}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 1, "pct": 6.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc186/arc186_b", "snap_from": 0, "snap_to": 1}]}], "label_dist": [{"label": "fix_bug", "count": 9, "pct": 56.2}, {"label": "rewrite", "count": 3, "pct": 18.8}, {"label": "optimize_perf", "count": 2, "pct": 12.5}, {"label": "implement", "count": 1, "pct": 6.2}, {"label": "infra_confusion", "count": 1, "pct": 6.2}], "n_labeled": 16, "nlp_rates": {"m3_existing": 62, "m1_efficiency": 31, "m2": 69, "m4": 50, "m8_inline": 25, "m9": 31, "m7": 12, "m1_correctness": 56, "m3_sample": 25, "m6": 19, "m8_lang": 6, "m3_own": 6}}, {"id": "m7", "source": "improved", "target": "improved", "color": "#2196F3", "count": 6, "out_pct": 7.1, "label": "6", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 5, "pct": 83.3, "examples": [{"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc389/abc389_g", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 11, "snap_to": 12}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 1, "snap_to": 2}]}, {"chain": ["run_sample", "time", "write"], "label": "run solution.py (sample) → time solution.py → write solution.py", "count": 1, "pct": 16.7, "examples": [{"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": 0, "snap_to": 1}]}], "label_dist": [{"label": "fix_bug", "count": 4, "pct": 66.7}, {"label": "optimize_perf", "count": 2, "pct": 33.3}], "n_labeled": 6, "nlp_rates": {"m1_correctness": 83, "m1_efficiency": 33, "m3_existing": 17, "m4": 33, "m2": 83, "m3_sample": 67}}, {"id": "m8", "source": "breakthrough", "target": "stay", "color": "#00BCD4", "count": 68, "out_pct": 18.8, "label": "68", "n_unique_chains": 12, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 48, "pct": 70.6, "examples": [{"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc302/abc302_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 11, "snap_to": 12}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 16, "snap_to": 17}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 23, "snap_to": 24}, {"problem": "gemma_4_26b_a4b_it/abc344/abc344_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc345/abc345_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc345/abc345_d", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc359/abc359_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc359/abc359_d", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc365/abc365_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc366/abc366_b", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc371/abc371_a", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc372/abc372_a", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc381/abc381_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc381/abc381_d", "snap_from": 14, "snap_to": 15}, {"problem": "gemma_4_26b_a4b_it/abc381/abc381_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc383/abc383_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc385/abc385_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc387/abc387_b", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc390/abc390_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc391/abc391_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc393/abc393_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc397/abc397_b", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc397/abc397_e", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_b", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 5, "snap_to": 6}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 7, "pct": 10.3, "examples": [{"problem": "gemma_4_26b_a4b_it/abc303/abc303_a", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc332/abc332_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc333/abc333_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc340/abc340_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc355/abc355_a", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc324/abc324_d", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_sample", "write"], "label": "run solution.py (sample) → write solution.py", "count": 2, "pct": 2.9, "examples": [{"problem": "claude-sonnet-4-6/arc193/arc193_a", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc325/abc325_b", "snap_from": 0, "snap_to": 1}]}, {"chain": ["time", "write"], "label": "time solution.py → write solution.py", "count": 2, "pct": 2.9, "examples": [{"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc358/abc358_e", "snap_from": 0, "snap_to": 1}]}, {"chain": ["test", "inspect", "write"], "label": "run ./test.sh → inspect test file → write solution.py", "count": 2, "pct": 2.9, "examples": [{"problem": "gemma_4_26b_a4b_it/abc392/abc392_b", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc392/abc392_b", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 1, "pct": 1.5, "examples": [{"problem": "claude-sonnet-4-6/abc392/abc392_b", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inline_script", "write"], "label": "inline_script → write solution.py", "count": 1, "pct": 1.5, "examples": [{"problem": "gpt-5.5/abc373/abc373_f", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inspect"], "label": "inspect test file", "count": 1, "pct": 1.5, "examples": [{"problem": "gpt-5.5/abc392/abc392_b", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_script", "write"], "label": "run helper script → write solution.py", "count": 1, "pct": 1.5, "examples": [{"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 11, "snap_to": 12}]}, {"chain": ["write_script", "write_script", "write"], "label": "write helper script → write helper script → write solution.py", "count": 1, "pct": 1.5, "examples": [{"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 3, "snap_to": 4}]}, {"chain": ["write_script", "write"], "label": "write helper script → write solution.py", "count": 1, "pct": 1.5, "examples": [{"problem": "gemma_4_26b_a4b_it/abc391/abc391_e", "snap_from": 6, "snap_to": 7}]}, {"chain": ["inspect", "inspect", "inspect", "wc", "write"], "label": "inspect test file → inspect test file → inspect test file → wc (output size) → write solution.py", "count": 1, "pct": 1.5, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc191/arc191_d", "snap_from": 0, "snap_to": 1}]}], "label_dist": [{"label": "optimize_perf", "count": 19, "pct": 27.9}, {"label": "implement", "count": 15, "pct": 22.1}, {"label": "cosmetic_refactor", "count": 14, "pct": 20.6}, {"label": "fix_bug", "count": 10, "pct": 14.7}, {"label": "rewrite", "count": 4, "pct": 5.9}, {"label": "verify", "count": 3, "pct": 4.4}, {"label": "edge_case", "count": 2, "pct": 2.9}, {"label": "execute_plan", "count": 1, "pct": 1.5}], "n_labeled": 68, "nlp_rates": {"m1_efficiency": 65, "m3_existing": 37, "m3_sample": 60, "m4": 26, "m8_inline": 3, "m9": 12, "m1_correctness": 81, "m2": 81, "m6": 60, "m7": 9, "m3_own": 3, "m8_lang": 3}}, {"id": "m9", "source": "stay", "target": "stay", "color": "#4CAF50", "count": 234, "out_pct": 77.5, "label": "234", "n_unique_chains": 8, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 209, "pct": 89.3, "examples": [{"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": 4, "snap_to": 5}, {"problem": "gpt-5.5/abc392/abc392_b", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 11, "snap_to": 12}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 11, "snap_to": 12}]}, {"chain": ["write_script", "write"], "label": "write helper script → write solution.py", "count": 12, "pct": 5.1, "examples": [{"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_e", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_e", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc391/abc391_e", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc392/abc392_b", "snap_from": 11, "snap_to": 12}, {"problem": "gemma_4_26b_a4b_it/abc392/abc392_b", "snap_from": 18, "snap_to": 19}, {"problem": "gemma_4_26b_a4b_it/abc393/abc393_e", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_d", "snap_from": 8, "snap_to": 9}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 6, "pct": 2.6, "examples": [{"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc372/abc372_a", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc383/abc383_c", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc383/abc383_c", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc383/abc383_c", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_d", "snap_from": 10, "snap_to": 11}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 3, "pct": 1.3, "examples": [{"problem": "claude-sonnet-4-6/abc392/abc392_b", "snap_from": 1, "snap_to": 2}, {"problem": "claude-sonnet-4-6/abc392/abc392_b", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc392/abc392_b", "snap_from": 1, "snap_to": 2}]}, {"chain": ["wc", "time", "write"], "label": "wc (output size) → time solution.py → write solution.py", "count": 1, "pct": 0.4, "examples": [{"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": 3, "snap_to": 4}]}, {"chain": ["test", "inline_script", "write"], "label": "run ./test.sh → inline_script → write solution.py", "count": 1, "pct": 0.4, "examples": [{"problem": "gemma_4_26b_a4b_it/abc358/abc358_e", "snap_from": 8, "snap_to": 9}]}, {"chain": ["test", "run_sample", "write"], "label": "run ./test.sh → run solution.py (sample) → write solution.py", "count": 1, "pct": 0.4, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc392/abc392_b", "snap_from": 1, "snap_to": 2}]}, {"chain": ["run_sample", "inspect", "wc", "inspect", "run_sample", "test", "run_sample", "write"], "label": "run solution.py (sample) → inspect test file → wc (output size) → inspect test file → run solution.py (sample) → run ./test.sh → run solution.py (sample) → write solution.py", "count": 1, "pct": 0.4, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc392/abc392_b", "snap_from": 2, "snap_to": 3}]}], "label_dist": [{"label": "cosmetic_refactor", "count": 79, "pct": 33.8}, {"label": "optimize_perf", "count": 42, "pct": 17.9}, {"label": "implement", "count": 35, "pct": 15.0}, {"label": "fix_bug", "count": 24, "pct": 10.3}, {"label": "verify", "count": 20, "pct": 8.5}, {"label": "revert", "count": 15, "pct": 6.4}, {"label": "edge_case", "count": 6, "pct": 2.6}, {"label": "explore", "count": 4, "pct": 1.7}, {"label": "debug_failure", "count": 4, "pct": 1.7}, {"label": "infra_confusion", "count": 4, "pct": 1.7}, {"label": "execute_plan", "count": 1, "pct": 0.4}], "n_labeled": 234, "nlp_rates": {"m1_efficiency": 79, "m3_existing": 29, "m4": 18, "m1_correctness": 86, "m2": 88, "m9": 8, "m8_inline": 2, "m6": 75, "m7": 11, "m3_sample": 69, "m3_own": 8, "m8_lang": 5}}, {"id": "m10", "source": "stay", "target": "end_submit", "color": "#4CAF50", "count": 45, "out_pct": 14.9, "label": "45", "n_unique_chains": 6, "breakdown": [{"chain": ["submit"], "label": "submit", "count": 32, "pct": 71.1, "examples": [{"problem": "claude-sonnet-4-6/abc373/abc373_f", "snap_from": 5, "snap_to": -1}, {"problem": "claude-sonnet-4-6/abc392/abc392_b", "snap_from": 3, "snap_to": -1}, {"problem": "gpt-5.5/abc373/abc373_f", "snap_from": 1, "snap_to": -1}, {"problem": "gpt-5.5/abc392/abc392_b", "snap_from": 2, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/1883/1883_B", "snap_from": 6, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc302/abc302_c", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc303/abc303_a", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_d", "snap_from": 6, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 9, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 26, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc332/abc332_c", "snap_from": 5, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc333/abc333_c", "snap_from": 5, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 9, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc340/abc340_c", "snap_from": 2, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc344/abc344_d", "snap_from": 5, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc355/abc355_a", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc358/abc358_e", "snap_from": 15, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc359/abc359_c", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc371/abc371_a", "snap_from": 3, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc372/abc372_a", "snap_from": 2, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 15, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc381/abc381_e", "snap_from": 10, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc383/abc383_c", "snap_from": 8, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc385/abc385_e", "snap_from": 11, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc387/abc387_b", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_e", "snap_from": 3, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc391/abc391_e", "snap_from": 9, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc393/abc393_e", "snap_from": 5, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc397/abc397_b", "snap_from": 2, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 6, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_d", "snap_from": 4, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc397/abc397_e", "snap_from": 3, "snap_to": -1}]}, {"chain": ["test", "submit"], "label": "run ./test.sh → submit", "count": 9, "pct": 20.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc310/abc310_b", "snap_from": 8, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc365/abc365_d", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_d", "snap_from": 7, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_b", "snap_from": 7, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc341/abc341_d", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc356/abc356_d", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_e", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/arc193/arc193_a", "snap_from": 1, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/arc194/arc194_b", "snap_from": 1, "snap_to": -1}]}, {"chain": ["run_sample", "test", "test", "submit"], "label": "run solution.py (sample) → run ./test.sh → run ./test.sh → submit", "count": 1, "pct": 2.2, "examples": [{"problem": "claude-sonnet-4-6/arc193/arc193_a", "snap_from": 2, "snap_to": -1}]}, {"chain": ["run_sample"], "label": "run solution.py (sample)", "count": 1, "pct": 2.2, "examples": [{"problem": "gemma_4_26b_a4b_it/abc345/abc345_d", "snap_from": 7, "snap_to": -1}]}, {"chain": ["run_sample", "write_script", "test", "submit"], "label": "run solution.py (sample) → write helper script → run ./test.sh → submit", "count": 1, "pct": 2.2, "examples": [{"problem": "gemma_4_26b_a4b_it/abc390/abc390_e", "snap_from": 3, "snap_to": -1}]}, {"chain": ["inspect", "inspect", "test", "submit"], "label": "inspect test file → inspect test file → run ./test.sh → submit", "count": 1, "pct": 2.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc191/arc191_d", "snap_from": 1, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 42, "pct": 95.5}, {"label": "infra_confusion", "count": 2, "pct": 4.5}], "n_labeled": 44, "nlp_rates": {"m3_existing": 31, "m3_sample": 9, "m1_correctness": 76, "m1_efficiency": 49, "m2": 62, "m6": 33, "m7": 2, "m9": 4, "m8_lang": 2, "m3_own": 2}}, {"id": "m11", "source": "improved", "target": "regressed", "color": "#2196F3", "count": 14, "out_pct": 16.7, "label": "14", "n_unique_chains": 6, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 7, "pct": 50.0, "examples": [{"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 12, "snap_to": 13}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 2, "snap_to": 3}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 2, "snap_to": 3}]}, {"chain": ["write", "write"], "label": "write solution.py → write solution.py", "count": 2, "pct": 14.3, "examples": [{"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 0, "snap_to": 1}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 2, "pct": 14.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": 0, "snap_to": 1}]}, {"chain": ["test", "inspect", "write_script", "inline_script", "write_script", "inline_script", "inline_script", "write"], "label": "run ./test.sh → inspect test file → write helper script → inline_script → write helper script → inline_script → inline_script → write solution.py", "count": 1, "pct": 7.1, "examples": [{"problem": "claude-sonnet-4-6/abc389/abc389_g", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inspect", "inspect", "inspect", "inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 7.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc337/abc337_e", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_sample", "run_sample", "write"], "label": "run solution.py (sample) → run solution.py (sample) → write solution.py", "count": 1, "pct": 7.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc390/abc390_e", "snap_from": 0, "snap_to": 1}]}], "label_dist": [{"label": "fix_bug", "count": 5, "pct": 35.7}, {"label": "rewrite", "count": 4, "pct": 28.6}, {"label": "optimize_perf", "count": 3, "pct": 21.4}, {"label": "implement", "count": 2, "pct": 14.3}], "n_labeled": 14, "nlp_rates": {"m1_correctness": 93, "m1_efficiency": 50, "m3_existing": 36, "m3_sample": 57, "m3_own": 29, "m4": 43, "m8_lang": 7, "m8_inline": 7, "m9": 50, "m2": 93, "m6": 36}}, {"id": "m12", "source": "regressed", "target": "improved", "color": "#F44336", "count": 8, "out_pct": 29.6, "label": "8", "n_unique_chains": 3, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 6, "pct": 75.0, "examples": [{"problem": "claude-sonnet-4-6/abc389/abc389_g", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 3, "snap_to": 4}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 1, "snap_to": 2}]}, {"chain": ["test", "inspect", "inspect", "write"], "label": "run ./test.sh → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 12.5, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc337/abc337_e", "snap_from": 1, "snap_to": 2}]}, {"chain": ["inspect", "inspect", "write"], "label": "inspect test file → inspect test file → write solution.py", "count": 1, "pct": 12.5, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 2, "snap_to": 3}]}], "label_dist": [{"label": "fix_bug", "count": 3, "pct": 37.5}, {"label": "implement", "count": 2, "pct": 25.0}, {"label": "rewrite", "count": 2, "pct": 25.0}, {"label": "optimize_perf", "count": 1, "pct": 12.5}], "n_labeled": 8, "nlp_rates": {"m1_efficiency": 38, "m2": 100, "m3_sample": 50, "m3_own": 38, "m4": 25, "m9": 50, "m1_correctness": 75, "m3_existing": 38, "m6": 38, "m7": 12}}, {"id": "m13", "source": "improved", "target": "stuck", "color": "#2196F3", "count": 24, "out_pct": 28.6, "label": "24", "n_unique_chains": 4, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 21, "pct": 87.5, "examples": [{"problem": "claude-sonnet-4-6/abc389/abc389_g", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc378/abc378_g", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc395/abc395_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/arc196/arc196_a", "snap_from": 1, "snap_to": 2}, {"problem": "qwen3p6_35b_a3b_fp8/abc345/abc345_d", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc385/abc385_e", "snap_from": 3, "snap_to": 4}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 5, "snap_to": 6}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 8, "snap_to": 9}, {"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/arc186/arc186_e", "snap_from": 0, "snap_to": 1}]}, {"chain": ["run_sample", "test", "write"], "label": "run solution.py (sample) → run ./test.sh → write solution.py", "count": 1, "pct": 4.2, "examples": [{"problem": "gpt-5.5/abc389/abc389_g", "snap_from": 1, "snap_to": 2}]}, {"chain": ["run_sample", "run_sample", "write"], "label": "run solution.py (sample) → run solution.py (sample) → write solution.py", "count": 1, "pct": 4.2, "examples": [{"problem": "gemma_4_26b_a4b_it/abc389/abc389_g", "snap_from": 3, "snap_to": 4}]}, {"chain": ["inspect", "inspect", "time", "run_sample", "run_sample", "write"], "label": "inspect test file → inspect test file → time solution.py → run solution.py (sample) → run solution.py (sample) → write solution.py", "count": 1, "pct": 4.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 4, "snap_to": 5}]}], "label_dist": [{"label": "cosmetic_refactor", "count": 7, "pct": 29.2}, {"label": "implement", "count": 5, "pct": 20.8}, {"label": "optimize_perf", "count": 3, "pct": 12.5}, {"label": "fix_bug", "count": 2, "pct": 8.3}, {"label": "debug_failure", "count": 2, "pct": 8.3}, {"label": "explore", "count": 2, "pct": 8.3}, {"label": "rewrite", "count": 2, "pct": 8.3}, {"label": "revert", "count": 1, "pct": 4.2}], "n_labeled": 24, "nlp_rates": {"m1_efficiency": 54, "m2": 71, "m3_sample": 58, "m3_own": 8, "m4": 33, "m9": 17, "m1_correctness": 79, "m3_existing": 17, "m6": 21, "m8_lang": 4, "m7": 4, "m8_inline": 4}}, {"id": "m14", "source": "improved", "target": "end_limit", "color": "#2196F3", "count": 13, "out_pct": 15.5, "label": "13", "n_unique_chains": 10, "breakdown": [{"chain": [], "label": "(no useful command)", "count": 4, "pct": 30.8, "examples": [{"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 11, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 10, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc192/arc192_b", "snap_from": 0, "snap_to": -1}, {"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 6, "snap_to": -1}]}, {"chain": ["inspect", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script"], "label": "inspect test file → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script", "count": 1, "pct": 7.7, "examples": [{"problem": "claude-sonnet-4-6/abc398/abc398_g", "snap_from": 0, "snap_to": -1}]}, {"chain": ["run_sample", "test", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script"], "label": "run solution.py (sample) → run ./test.sh → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script", "count": 1, "pct": 7.7, "examples": [{"problem": "claude-sonnet-4-6/abc400/abc400_g", "snap_from": 0, "snap_to": -1}]}, {"chain": ["inspect", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script", "inline_script"], "label": "inspect test file → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script → inline_script", "count": 1, "pct": 7.7, "examples": [{"problem": "claude-sonnet-4-6/arc192/arc192_b", "snap_from": 0, "snap_to": -1}]}, {"chain": ["run_sample", "test"], "label": "run solution.py (sample) → run ./test.sh", "count": 1, "pct": 7.7, "examples": [{"problem": "gemma_4_26b_a4b_it/arc196/arc196_d", "snap_from": 7, "snap_to": -1}]}, {"chain": ["inspect", "wc"], "label": "inspect test file → wc (output size)", "count": 1, "pct": 7.7, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc337/abc337_e", "snap_from": 2, "snap_to": -1}]}, {"chain": ["test", "inspect", "inspect"], "label": "run ./test.sh → inspect test file → inspect test file", "count": 1, "pct": 7.7, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 12, "snap_to": -1}]}, {"chain": ["run_sample", "write_script", "write_script", "inspect", "inspect", "write_script", "write_script", "inspect", "inspect", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script"], "label": "run solution.py (sample) → write helper script → write helper script → inspect test file → inspect test file → write helper script → write helper script → inspect test file → inspect test file → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script", "count": 1, "pct": 7.7, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc184/arc184_d", "snap_from": 1, "snap_to": -1}]}, {"chain": ["test", "inspect", "inspect", "inspect", "inspect", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script"], "label": "run ./test.sh → inspect test file → inspect test file → inspect test file → inspect test file → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script", "count": 1, "pct": 7.7, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc192/arc192_b", "snap_from": 0, "snap_to": -1}]}, {"chain": ["write_script", "test", "test", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script"], "label": "write helper script → run ./test.sh → run ./test.sh → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script", "count": 1, "pct": 7.7, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc194/arc194_d", "snap_from": 2, "snap_to": -1}]}], "label_dist": [{"label": "debug_failure", "count": 5, "pct": 38.5}, {"label": "verify", "count": 4, "pct": 30.8}, {"label": "explore", "count": 4, "pct": 30.8}], "n_labeled": 13, "nlp_rates": {"m1_correctness": 54, "m3_existing": 69, "m4": 8, "m8_lang": 31, "m9": 62, "m1_efficiency": 38, "m8_inline": 15, "m3_sample": 8, "m2": 38, "m6": 31}}, {"id": "m15", "source": "stuck", "target": "breakthrough", "color": "#FF9800", "count": 21, "out_pct": 13.7, "label": "21", "n_unique_chains": 8, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 14, "pct": 66.7, "examples": [{"problem": "claude-sonnet-4-6/arc193/arc193_a", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 15, "snap_to": 16}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 22, "snap_to": 23}, {"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc359/abc359_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc384/abc384_c", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/abc395/abc395_c", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_b", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc351/abc351_a", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inline_script", "inline_script"], "label": "inline_script → inline_script", "count": 1, "pct": 4.8, "examples": [{"problem": "gpt-5.5/abc389/abc389_g", "snap_from": 2, "snap_to": 3}]}, {"chain": [], "label": "(no useful command)", "count": 1, "pct": 4.8, "examples": [{"problem": "gpt-5.5/arc194/arc194_a", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inline_script", "inline_script", "inline_script", "inline_script", "write"], "label": "inline_script → inline_script → inline_script → inline_script → write solution.py", "count": 1, "pct": 4.8, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc324/abc324_d", "snap_from": 7, "snap_to": 8}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 1, "pct": 4.8, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc345/abc345_d", "snap_from": 1, "snap_to": 2}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 1, "pct": 4.8, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc354/abc354_d", "snap_from": 0, "snap_to": 1}]}, {"chain": ["test", "inspect", "write"], "label": "run ./test.sh → inspect test file → write solution.py", "count": 1, "pct": 4.8, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc365/abc365_d", "snap_from": 1, "snap_to": 2}]}, {"chain": ["inspect", "run_sample", "write_script", "write_script", "write"], "label": "inspect test file → run solution.py (sample) → write helper script → write helper script → write solution.py", "count": 1, "pct": 4.8, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc385/abc385_e", "snap_from": 4, "snap_to": 5}]}], "label_dist": [{"label": "fix_bug", "count": 8, "pct": 38.1}, {"label": "execute_plan", "count": 4, "pct": 19.0}, {"label": "optimize_perf", "count": 4, "pct": 19.0}, {"label": "implement", "count": 2, "pct": 9.5}, {"label": "rewrite", "count": 2, "pct": 9.5}, {"label": "infra_confusion", "count": 1, "pct": 4.8}], "n_labeled": 21, "nlp_rates": {"m3_sample": 52, "m4": 48, "m9": 14, "m1_efficiency": 33, "m3_existing": 29, "m8_inline": 10, "m1_correctness": 57, "m2": 62, "m6": 19, "m7": 5, "m8_lang": 5}}, {"id": "m16", "source": "improved", "target": "broke", "color": "#2196F3", "count": 5, "out_pct": 6.0, "label": "5", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 4, "pct": 80.0, "examples": [{"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_a", "snap_from": 4, "snap_to": 5}, {"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 3, "snap_to": 4}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 2, "snap_to": 3}]}, {"chain": ["inspect", "inspect", "write"], "label": "inspect test file → inspect test file → write solution.py", "count": 1, "pct": 20.0, "examples": [{"problem": "claude-sonnet-4-6/arc196/arc196_a", "snap_from": 0, "snap_to": 1}]}], "label_dist": [{"label": "explore", "count": 3, "pct": 60.0}, {"label": "implement", "count": 1, "pct": 20.0}, {"label": "debug_failure", "count": 1, "pct": 20.0}], "n_labeled": 5, "nlp_rates": {"m1_correctness": 100, "m3_existing": 20, "m4": 40, "m1_efficiency": 60, "m2": 80, "m6": 20}}, {"id": "m17", "source": "broke", "target": "end_limit", "color": "#7B0000", "count": 1, "out_pct": 4.0, "label": "1", "n_unique_chains": 1, "breakdown": [{"chain": ["write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "write_script", "run_script", "write_script", "write_script"], "label": "write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → write helper script → run helper script → write helper script → write helper script", "count": 1, "pct": 100.0, "examples": [{"problem": "claude-sonnet-4-6/arc196/arc196_a", "snap_from": 1, "snap_to": -1}]}], "label_dist": [{"label": "explore", "count": 1, "pct": 100.0}], "n_labeled": 1, "nlp_rates": {"m3_existing": 100, "m4": 100, "m8_lang": 100, "m9": 100}}, {"id": "m18", "source": "stuck", "target": "improved", "color": "#FF9800", "count": 28, "out_pct": 18.3, "label": "28", "n_unique_chains": 4, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 24, "pct": 85.7, "examples": [{"problem": "claude-sonnet-4-6/arc196/arc196_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/abc378/abc378_g", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc389/abc389_g", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc398/abc398_g", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 18, "snap_to": 19}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_a", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc196/arc196_a", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/arc196/arc196_d", "snap_from": 6, "snap_to": 7}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 11, "snap_to": 12}, {"problem": "qwen3p6_35b_a3b_fp8/abc391/abc391_e", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 5, "snap_to": 6}, {"problem": "qwen3p6_35b_a3b_fp8/arc184/arc184_d", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/arc194/arc194_d", "snap_from": 1, "snap_to": 2}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 0, "snap_to": 1}]}, {"chain": ["write_script", "write"], "label": "write helper script → write solution.py", "count": 2, "pct": 7.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc385/abc385_e", "snap_from": 2, "snap_to": 3}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 1, "snap_to": 2}]}, {"chain": ["inline_script"], "label": "inline_script", "count": 1, "pct": 3.6, "examples": [{"problem": "gpt-5.5/abc389/abc389_g", "snap_from": 0, "snap_to": 1}]}, {"chain": ["inline_script", "write"], "label": "inline_script → write solution.py", "count": 1, "pct": 3.6, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc390/abc390_e", "snap_from": 3, "snap_to": 4}]}], "label_dist": [{"label": "fix_bug", "count": 10, "pct": 35.7}, {"label": "implement", "count": 6, "pct": 21.4}, {"label": "execute_plan", "count": 6, "pct": 21.4}, {"label": "rewrite", "count": 5, "pct": 17.9}, {"label": "explore", "count": 1, "pct": 3.6}], "n_labeled": 28, "nlp_rates": {"m1_correctness": 71, "m3_existing": 7, "m8_inline": 7, "m9": 11, "m3_sample": 54, "m4": 54, "m2": 75, "m1_efficiency": 29, "m3_own": 11, "m6": 14, "m8_lang": 7}}, {"id": "m19", "source": "stuck", "target": "broke", "color": "#FF9800", "count": 6, "out_pct": 3.9, "label": "6", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 6, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc378/abc378_g", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": 7, "snap_to": 8}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 6, "snap_to": 7}]}], "label_dist": [{"label": "explore", "count": 3, "pct": 50.0}, {"label": "rewrite", "count": 1, "pct": 16.7}, {"label": "debug_failure", "count": 1, "pct": 16.7}, {"label": "implement", "count": 1, "pct": 16.7}], "n_labeled": 6, "nlp_rates": {"m1_correctness": 67, "m1_efficiency": 67, "m2": 100, "m3_sample": 67, "m4": 67, "m9": 17}}, {"id": "m20", "source": "broke", "target": "improved", "color": "#7B0000", "count": 7, "out_pct": 28.0, "label": "7", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 6, "pct": 85.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": 8, "snap_to": 9}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 4, "snap_to": 5}, {"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 7, "snap_to": 8}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 1, "pct": 14.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 3, "snap_to": 4}]}], "label_dist": [{"label": "fix_bug", "count": 3, "pct": 42.9}, {"label": "rewrite", "count": 2, "pct": 28.6}, {"label": "implement", "count": 1, "pct": 14.3}, {"label": "execute_plan", "count": 1, "pct": 14.3}], "n_labeled": 7, "nlp_rates": {"m1_correctness": 71, "m2": 86, "m3_sample": 86, "m4": 14, "m6": 14, "m1_efficiency": 14, "m3_existing": 14}}, {"id": "m21", "source": "improved", "target": "churn", "color": "#2196F3", "count": 5, "out_pct": 6.0, "label": "5", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 5, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 4, "snap_to": 5}, {"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/abc391/abc391_e", "snap_from": 1, "snap_to": 2}, {"problem": "qwen3p6_35b_a3b_fp8/abc398/abc398_g", "snap_from": 0, "snap_to": 1}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": 3, "snap_to": 4}]}], "label_dist": [{"label": "rewrite", "count": 4, "pct": 80.0}, {"label": "fix_bug", "count": 1, "pct": 20.0}], "n_labeled": 5, "nlp_rates": {"m1_correctness": 80, "m1_efficiency": 80, "m2": 100, "m3_sample": 20, "m4": 40, "m3_existing": 20}}, {"id": "m22", "source": "churn", "target": "broke", "color": "#9C27B0", "count": 2, "out_pct": 8.0, "label": "2", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 2, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 8, "snap_to": 9}]}], "label_dist": [{"label": "rewrite", "count": 2, "pct": 100.0}], "n_labeled": 2, "nlp_rates": {"m1_correctness": 50, "m2": 100, "m3_sample": 50, "m4": 50, "m9": 50}}, {"id": "m23", "source": "broke", "target": "stuck", "color": "#7B0000", "count": 11, "out_pct": 44.0, "label": "11", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 10, "pct": 90.9, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 14, "snap_to": 15}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 21, "snap_to": 22}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc378/abc378_g", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_a", "snap_from": 5, "snap_to": 6}, {"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 4, "snap_to": 5}]}, {"chain": ["run_sample", "write"], "label": "run solution.py (sample) → write solution.py", "count": 1, "pct": 9.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc385/abc385_e", "snap_from": 1, "snap_to": 2}]}], "label_dist": [{"label": "explore", "count": 5, "pct": 45.5}, {"label": "debug_failure", "count": 4, "pct": 36.4}, {"label": "optimize_perf", "count": 1, "pct": 9.1}, {"label": "execute_plan", "count": 1, "pct": 9.1}], "n_labeled": 11, "nlp_rates": {"m1_correctness": 55, "m2": 64, "m4": 73, "m1_efficiency": 36, "m3_existing": 9, "m9": 9, "m6": 9}}, {"id": "m24", "source": "stuck", "target": "end_submit", "color": "#FF9800", "count": 7, "out_pct": 4.6, "label": "7", "n_unique_chains": 3, "breakdown": [{"chain": ["submit"], "label": "submit", "count": 5, "pct": 71.4, "examples": [{"problem": "gemma_4_26b_a4b_it/abc301/abc301_f", "snap_from": 10, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc382/abc382_g", "snap_from": 4, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc389/abc389_g", "snap_from": 4, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 9, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc195/arc195_d", "snap_from": 11, "snap_to": -1}]}, {"chain": ["write_script", "submit"], "label": "write helper script → submit", "count": 1, "pct": 14.3, "examples": [{"problem": "gemma_4_26b_a4b_it/abc337/abc337_e", "snap_from": 0, "snap_to": -1}]}, {"chain": [], "label": "(no useful command)", "count": 1, "pct": 14.3, "examples": [{"problem": "gemma_4_26b_a4b_it/arc196/arc196_a", "snap_from": 2, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 6, "pct": 100.0}], "n_labeled": 6, "nlp_rates": {"m1_correctness": 86, "m1_efficiency": 71, "m2": 71, "m7": 14, "m8_lang": 14, "m6": 29, "m3_sample": 14, "m9": 14}}, {"id": "m25", "source": "breakthrough", "target": "regressed", "color": "#00BCD4", "count": 6, "out_pct": 1.7, "label": "6", "n_unique_chains": 3, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 4, "pct": 66.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 0, "snap_to": 1}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 7, "snap_to": 8}, {"problem": "qwen3p6_35b_a3b_fp8/abc345/abc345_d", "snap_from": 2, "snap_to": 3}, {"problem": "qwen3p6_35b_a3b_fp8/abc397/abc397_e", "snap_from": 0, "snap_to": 1}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 1, "pct": 16.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 5, "snap_to": 6}]}, {"chain": ["inspect", "inspect", "write"], "label": "inspect test file → inspect test file → write solution.py", "count": 1, "pct": 16.7, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc324/abc324_d", "snap_from": 5, "snap_to": 6}]}], "label_dist": [{"label": "implement", "count": 3, "pct": 50.0}, {"label": "optimize_perf", "count": 1, "pct": 16.7}, {"label": "rewrite", "count": 1, "pct": 16.7}, {"label": "fix_bug", "count": 1, "pct": 16.7}], "n_labeled": 6, "nlp_rates": {"m1_correctness": 83, "m2": 100, "m3_sample": 50, "m4": 33, "m6": 50, "m1_efficiency": 83, "m3_existing": 33, "m7": 17, "m9": 17}}, {"id": "m26", "source": "regressed", "target": "stuck", "color": "#F44336", "count": 7, "out_pct": 25.9, "label": "7", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 6, "pct": 85.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc308/abc308_e", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 8, "snap_to": 9}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 1, "snap_to": 2}, {"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 7, "snap_to": 8}, {"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 13, "snap_to": 14}]}, {"chain": ["inline_script", "write"], "label": "inline_script → write solution.py", "count": 1, "pct": 14.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc324/abc324_d", "snap_from": 6, "snap_to": 7}]}], "label_dist": [{"label": "fix_bug", "count": 2, "pct": 28.6}, {"label": "rewrite", "count": 1, "pct": 14.3}, {"label": "optimize_perf", "count": 1, "pct": 14.3}, {"label": "cosmetic_refactor", "count": 1, "pct": 14.3}, {"label": "implement", "count": 1, "pct": 14.3}, {"label": "debug_failure", "count": 1, "pct": 14.3}], "n_labeled": 7, "nlp_rates": {"m1_correctness": 71, "m2": 86, "m3_sample": 86, "m4": 14, "m6": 57, "m1_efficiency": 57, "m3_own": 14, "m9": 14, "m8_inline": 14}}, {"id": "m27", "source": "stay", "target": "broke", "color": "#4CAF50", "count": 9, "out_pct": 3.0, "label": "9", "n_unique_chains": 3, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 6, "pct": 66.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 5, "snap_to": 6}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 13, "snap_to": 14}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 20, "snap_to": 21}, {"problem": "gemma_4_26b_a4b_it/abc345/abc345_d", "snap_from": 3, "snap_to": 4}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 9, "snap_to": 10}]}, {"chain": ["write_script", "write"], "label": "write helper script → write solution.py", "count": 2, "pct": 22.2, "examples": [{"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 9, "snap_to": 10}, {"problem": "gemma_4_26b_a4b_it/abc391/abc391_e", "snap_from": 4, "snap_to": 5}]}, {"chain": ["write", "write"], "label": "write solution.py → write solution.py", "count": 1, "pct": 11.1, "examples": [{"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 1, "snap_to": 2}]}], "label_dist": [{"label": "explore", "count": 7, "pct": 77.8}, {"label": "infra_confusion", "count": 1, "pct": 11.1}, {"label": "rewrite", "count": 1, "pct": 11.1}], "n_labeled": 9, "nlp_rates": {"m1_correctness": 78, "m1_efficiency": 78, "m2": 78, "m3_own": 11, "m6": 67, "m4": 89, "m8_lang": 22, "m9": 22, "m3_existing": 11}}, {"id": "m28", "source": "broke", "target": "breakthrough", "color": "#7B0000", "count": 6, "out_pct": 24.0, "label": "6", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 6, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 6, "snap_to": 7}, {"problem": "gemma_4_26b_a4b_it/abc324/abc324_d", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/abc345/abc345_d", "snap_from": 4, "snap_to": 5}, {"problem": "gemma_4_26b_a4b_it/abc373/abc373_f", "snap_from": 10, "snap_to": 11}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 2, "snap_to": 3}, {"problem": "gemma_4_26b_a4b_it/abc391/abc391_e", "snap_from": 5, "snap_to": 6}]}], "label_dist": [{"label": "rewrite", "count": 2, "pct": 33.3}, {"label": "cosmetic_refactor", "count": 1, "pct": 16.7}, {"label": "fix_bug", "count": 1, "pct": 16.7}, {"label": "implement", "count": 1, "pct": 16.7}, {"label": "execute_plan", "count": 1, "pct": 16.7}], "n_labeled": 6, "nlp_rates": {"m1_correctness": 50, "m1_efficiency": 50, "m2": 50, "m3_sample": 83, "m6": 17, "m4": 50}}, {"id": "m29", "source": "stay", "target": "end_limit", "color": "#4CAF50", "count": 11, "out_pct": 3.6, "label": "11", "n_unique_chains": 4, "breakdown": [{"chain": [], "label": "(no useful command)", "count": 8, "pct": 72.7, "examples": [{"problem": "gemma_4_26b_a4b_it/abc310/abc310_f", "snap_from": 12, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc312/abc312_d", "snap_from": 18, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc325/abc325_b", "snap_from": 2, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc366/abc366_b", "snap_from": 1, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc381/abc381_d", "snap_from": 17, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc397/abc397_e", "snap_from": 11, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/abc400/abc400_d", "snap_from": 11, "snap_to": -1}, {"problem": "gemma_4_26b_a4b_it/arc194/arc194_e", "snap_from": 11, "snap_to": -1}]}, {"chain": ["test"], "label": "run ./test.sh", "count": 1, "pct": 9.1, "examples": [{"problem": "gemma_4_26b_a4b_it/abc359/abc359_d", "snap_from": 14, "snap_to": -1}]}, {"chain": ["write_script"], "label": "write helper script", "count": 1, "pct": 9.1, "examples": [{"problem": "gemma_4_26b_a4b_it/abc392/abc392_b", "snap_from": 21, "snap_to": -1}]}, {"chain": ["inspect", "inspect", "inspect", "test", "run_sample", "inline_script", "run_sample", "test", "run_sample", "run_sample", "inline_script", "inline_script"], "label": "inspect test file → inspect test file → inspect test file → run ./test.sh → run solution.py (sample) → inline_script → run solution.py (sample) → run ./test.sh → run solution.py (sample) → run solution.py (sample) → inline_script → inline_script", "count": 1, "pct": 9.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc392/abc392_b", "snap_from": 3, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 10, "pct": 83.3}, {"label": "debug_failure", "count": 2, "pct": 16.7}], "n_labeled": 12, "nlp_rates": {"m1_correctness": 18, "m1_efficiency": 9, "m2": 18, "m3_existing": 27, "m8_lang": 9, "m6": 9, "m8_inline": 9, "m9": 9}}, {"id": "m30", "source": "regressed", "target": "breakthrough", "color": "#F44336", "count": 4, "out_pct": 14.8, "label": "4", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 4, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc334/abc334_c", "snap_from": 6, "snap_to": 7}, {"problem": "qwen3p6_35b_a3b_fp8/abc324/abc324_d", "snap_from": 4, "snap_to": 5}, {"problem": "qwen3p6_35b_a3b_fp8/abc345/abc345_d", "snap_from": 3, "snap_to": 4}, {"problem": "qwen3p6_35b_a3b_fp8/abc397/abc397_e", "snap_from": 1, "snap_to": 2}]}], "label_dist": [{"label": "fix_bug", "count": 2, "pct": 50.0}, {"label": "optimize_perf", "count": 1, "pct": 25.0}, {"label": "implement", "count": 1, "pct": 25.0}], "n_labeled": 4, "nlp_rates": {"m1_correctness": 50, "m1_efficiency": 75, "m2": 100, "m3_sample": 25, "m6": 50, "m7": 25}}, {"id": "m31", "source": "stay", "target": "regressed", "color": "#4CAF50", "count": 3, "out_pct": 1.0, "label": "3", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 3, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc381/abc381_d", "snap_from": 11, "snap_to": 12}, {"problem": "gemma_4_26b_a4b_it/abc388/abc388_f", "snap_from": 5, "snap_to": 6}, {"problem": "qwen3p6_35b_a3b_fp8/abc324/abc324_d", "snap_from": 3, "snap_to": 4}]}], "label_dist": [{"label": "optimize_perf", "count": 2, "pct": 66.7}, {"label": "cosmetic_refactor", "count": 1, "pct": 33.3}], "n_labeled": 3, "nlp_rates": {"m1_correctness": 67, "m1_efficiency": 100, "m2": 100, "m3_sample": 67, "m6": 33, "m7": 33}}, {"id": "m32", "source": "regressed", "target": "churn", "color": "#F44336", "count": 5, "out_pct": 18.5, "label": "5", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 5, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc381/abc381_d", "snap_from": 12, "snap_to": 13}, {"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 6, "snap_to": 7}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 3, "snap_to": 4}, {"problem": "qwen3p6_35b_a3b_fp8/abc390/abc390_e", "snap_from": 1, "snap_to": 2}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": 1, "snap_to": 2}]}], "label_dist": [{"label": "fix_bug", "count": 2, "pct": 40.0}, {"label": "edge_case", "count": 1, "pct": 20.0}, {"label": "optimize_perf", "count": 1, "pct": 20.0}, {"label": "rewrite", "count": 1, "pct": 20.0}], "n_labeled": 5, "nlp_rates": {"m1_correctness": 60, "m1_efficiency": 80, "m2": 100, "m3_sample": 40, "m6": 40, "m4": 20, "m9": 20}}, {"id": "m33", "source": "churn", "target": "breakthrough", "color": "#9C27B0", "count": 2, "out_pct": 8.0, "label": "2", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 1, "pct": 50.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc381/abc381_d", "snap_from": 13, "snap_to": 14}]}, {"chain": ["inline_script", "write"], "label": "inline_script → write solution.py", "count": 1, "pct": 50.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc391/abc391_e", "snap_from": 4, "snap_to": 5}]}], "label_dist": [{"label": "revert", "count": 1, "pct": 50.0}, {"label": "fix_bug", "count": 1, "pct": 50.0}], "n_labeled": 2, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 50, "m2": 100, "m3_sample": 50, "m6": 100, "m7": 50, "m4": 50, "m8_inline": 50}}, {"id": "m34", "source": "breakthrough", "target": "broke", "color": "#00BCD4", "count": 2, "out_pct": 0.6, "label": "2", "n_unique_chains": 2, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 1, "pct": 50.0, "examples": [{"problem": "gemma_4_26b_a4b_it/abc400/abc400_g", "snap_from": 3, "snap_to": 4}]}, {"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 1, "pct": 50.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc385/abc385_e", "snap_from": 0, "snap_to": 1}]}], "label_dist": [{"label": "explore", "count": 1, "pct": 50.0}, {"label": "implement", "count": 1, "pct": 50.0}], "n_labeled": 2, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 100, "m2": 100, "m4": 50, "m3_existing": 50, "m6": 50, "m7": 50}}, {"id": "m35", "source": "regressed", "target": "regressed", "color": "#F44336", "count": 1, "out_pct": 3.7, "label": "1", "n_unique_chains": 1, "breakdown": [{"chain": ["run_sample", "write"], "label": "run solution.py (sample) → write solution.py", "count": 1, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/arc184/arc184_d", "snap_from": 4, "snap_to": 5}]}], "label_dist": [{"label": "rewrite", "count": 1, "pct": 100.0}], "n_labeled": 1, "nlp_rates": {"m1_correctness": 100, "m2": 100, "m3_sample": 100, "m9": 100}}, {"id": "m36", "source": "churn", "target": "churn", "color": "#9C27B0", "count": 11, "out_pct": 44.0, "label": "11", "n_unique_chains": 4, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 7, "pct": 63.6, "examples": [{"problem": "gemma_4_26b_a4b_it/arc186/arc186_e", "snap_from": 7, "snap_to": 8}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 4, "snap_to": 5}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 5, "snap_to": 6}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 6, "snap_to": 7}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 9, "snap_to": 10}, {"problem": "qwen3p6_35b_a3b_fp8/abc391/abc391_e", "snap_from": 3, "snap_to": 4}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 6, "snap_to": 7}]}, {"chain": ["inline_script", "write"], "label": "inline_script → write solution.py", "count": 2, "pct": 18.2, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 7, "snap_to": 8}, {"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 8, "snap_to": 9}]}, {"chain": ["test", "inspect", "inspect", "inspect", "inspect", "write"], "label": "run ./test.sh → inspect test file → inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 9.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 3, "snap_to": 4}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 1, "pct": 9.1, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc391/abc391_e", "snap_from": 2, "snap_to": 3}]}], "label_dist": [{"label": "rewrite", "count": 6, "pct": 54.5}, {"label": "debug_failure", "count": 2, "pct": 18.2}, {"label": "fix_bug", "count": 2, "pct": 18.2}, {"label": "implement", "count": 1, "pct": 9.1}], "n_labeled": 11, "nlp_rates": {"m2": 100, "m3_sample": 9, "m1_correctness": 91, "m1_efficiency": 73, "m6": 55, "m7": 9, "m8_inline": 18, "m3_existing": 27, "m9": 18, "m4": 18}}, {"id": "m37", "source": "stuck", "target": "regressed", "color": "#FF9800", "count": 3, "out_pct": 2.0, "label": "3", "n_unique_chains": 3, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 1, "pct": 33.3, "examples": [{"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 9, "snap_to": 10}]}, {"chain": ["test", "write"], "label": "run ./test.sh → write solution.py", "count": 1, "pct": 33.3, "examples": [{"problem": "gemma_4_26b_a4b_it/arc194/arc194_b", "snap_from": 2, "snap_to": 3}]}, {"chain": ["run_sample", "inspect", "write"], "label": "run solution.py (sample) → inspect test file → write solution.py", "count": 1, "pct": 33.3, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc400/abc400_d", "snap_from": 1, "snap_to": 2}]}], "label_dist": [{"label": "implement", "count": 2, "pct": 66.7}, {"label": "fix_bug", "count": 1, "pct": 33.3}], "n_labeled": 3, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 67, "m2": 100, "m3_sample": 33, "m4": 67, "m6": 67, "m3_existing": 67, "m9": 67}}, {"id": "m38", "source": "regressed", "target": "end_limit", "color": "#F44336", "count": 1, "out_pct": 3.7, "label": "1", "n_unique_chains": 1, "breakdown": [{"chain": [], "label": "(no useful command)", "count": 1, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/arc191/arc191_d", "snap_from": 10, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 1, "pct": 100.0}], "n_labeled": 1, "nlp_rates": {}}, {"id": "m39", "source": "improved", "target": "end_submit", "color": "#2196F3", "count": 1, "out_pct": 1.2, "label": "1", "n_unique_chains": 1, "breakdown": [{"chain": ["submit"], "label": "submit", "count": 1, "pct": 100.0, "examples": [{"problem": "gemma_4_26b_a4b_it/arc193/arc193_a", "snap_from": 19, "snap_to": -1}]}], "label_dist": [{"label": "verify", "count": 1, "pct": 100.0}], "n_labeled": 1, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 100, "m2": 100}}, {"id": "m40", "source": "churn", "target": "end_limit", "color": "#9C27B0", "count": 4, "out_pct": 16.0, "label": "4", "n_unique_chains": 4, "breakdown": [{"chain": ["test", "test", "inspect"], "label": "run ./test.sh → run ./test.sh → inspect test file", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc373/abc373_f", "snap_from": 10, "snap_to": -1}]}, {"chain": ["inspect", "write_script", "write_script", "write_script"], "label": "inspect test file → write helper script → write helper script → write helper script", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc398/abc398_g", "snap_from": 4, "snap_to": -1}]}, {"chain": ["inspect"], "label": "inspect test file", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc186/arc186_e", "snap_from": 2, "snap_to": -1}]}, {"chain": [], "label": "(no useful command)", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 7, "snap_to": -1}]}], "label_dist": [{"label": "explore", "count": 2, "pct": 50.0}, {"label": "verify", "count": 2, "pct": 50.0}], "n_labeled": 4, "nlp_rates": {"m1_correctness": 75, "m1_efficiency": 50, "m2": 50, "m3_existing": 75, "m6": 25, "m7": 25, "m9": 25, "m8_lang": 25}}, {"id": "m41", "source": "churn", "target": "stuck", "color": "#9C27B0", "count": 5, "out_pct": 20.0, "label": "5", "n_unique_chains": 2, "breakdown": [{"chain": ["inspect", "write"], "label": "inspect test file → write solution.py", "count": 3, "pct": 60.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 4, "snap_to": 5}, {"problem": "qwen3p6_35b_a3b_fp8/abc390/abc390_e", "snap_from": 2, "snap_to": 3}, {"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": 4, "snap_to": 5}]}, {"chain": ["write"], "label": "write solution.py", "count": 2, "pct": 40.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 1, "snap_to": 2}, {"problem": "qwen3p6_35b_a3b_fp8/abc398/abc398_g", "snap_from": 1, "snap_to": 2}]}], "label_dist": [{"label": "implement", "count": 2, "pct": 40.0}, {"label": "fix_bug", "count": 2, "pct": 40.0}, {"label": "rewrite", "count": 1, "pct": 20.0}], "n_labeled": 5, "nlp_rates": {"m1_efficiency": 80, "m2": 100, "m1_correctness": 80, "m3_existing": 60, "m9": 20, "m6": 20}}, {"id": "m42", "source": "stuck", "target": "churn", "color": "#FF9800", "count": 4, "out_pct": 2.6, "label": "4", "n_unique_chains": 4, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc388/abc388_f", "snap_from": 2, "snap_to": 3}]}, {"chain": ["inspect", "inspect", "inspect", "write"], "label": "inspect test file → inspect test file → inspect test file → write solution.py", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc398/abc398_g", "snap_from": 3, "snap_to": 4}]}, {"chain": ["write_script", "write_script", "write"], "label": "write helper script → write helper script → write solution.py", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc186/arc186_e", "snap_from": 1, "snap_to": 2}]}, {"chain": ["inspect", "write_script", "write"], "label": "inspect test file → write helper script → write solution.py", "count": 1, "pct": 25.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_a", "snap_from": 5, "snap_to": 6}]}], "label_dist": [{"label": "fix_bug", "count": 2, "pct": 50.0}, {"label": "rewrite", "count": 1, "pct": 25.0}, {"label": "implement", "count": 1, "pct": 25.0}], "n_labeled": 4, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 75, "m2": 100, "m3_existing": 75, "m6": 50, "m8_lang": 50}}, {"id": "m43", "source": "regressed", "target": "broke", "color": "#F44336", "count": 1, "out_pct": 3.7, "label": "1", "n_unique_chains": 1, "breakdown": [{"chain": ["write"], "label": "write solution.py", "count": 1, "pct": 100.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/abc389/abc389_g", "snap_from": 3, "snap_to": 4}]}], "label_dist": [{"label": "optimize_perf", "count": 1, "pct": 100.0}], "n_labeled": 1, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 100, "m2": 100}}, {"id": "m44", "source": "churn", "target": "improved", "color": "#9C27B0", "count": 1, "out_pct": 4.0, "label": "1", "n_unique_chains": 1, "breakdown": [{"chain": ["run_sample", "write"], "label": "run solution.py (sample) → write solution.py", "count": 1, "pct": 100.0, "examples": [{"problem": "qwen3p6_35b_a3b_fp8/arc196/arc196_d", "snap_from": 2, "snap_to": 3}]}], "label_dist": [{"label": "fix_bug", "count": 1, "pct": 100.0}], "n_labeled": 1, "nlp_rates": {"m1_correctness": 100, "m1_efficiency": 100, "m2": 100, "m3_existing": 100, "m6": 100}}], "prob_rows": [{"src": "start", "total": 381, "outs": [{"tgt": "breakthrough", "count": 313, "pct": 82.2}, {"tgt": "stuck", "count": 34, "pct": 8.9}, {"tgt": "improved", "count": 34, "pct": 8.9}]}, {"src": "improved", "total": 84, "outs": [{"tgt": "stuck", "count": 24, "pct": 28.6}, {"tgt": "breakthrough", "count": 16, "pct": 19.0}, {"tgt": "regressed", "count": 14, "pct": 16.7}, {"tgt": "end_limit", "count": 13, "pct": 15.5}, {"tgt": "improved", "count": 6, "pct": 7.1}, {"tgt": "broke", "count": 5, "pct": 6.0}, {"tgt": "churn", "count": 5, "pct": 6.0}, {"tgt": "end_submit", "count": 1, "pct": 1.2}]}, {"src": "churn", "total": 25, "outs": [{"tgt": "churn", "count": 11, "pct": 44.0}, {"tgt": "stuck", "count": 5, "pct": 20.0}, {"tgt": "end_limit", "count": 4, "pct": 16.0}, {"tgt": "broke", "count": 2, "pct": 8.0}, {"tgt": "breakthrough", "count": 2, "pct": 8.0}, {"tgt": "improved", "count": 1, "pct": 4.0}]}, {"src": "stuck", "total": 153, "outs": [{"tgt": "stuck", "count": 72, "pct": 47.1}, {"tgt": "improved", "count": 28, "pct": 18.3}, {"tgt": "breakthrough", "count": 21, "pct": 13.7}, {"tgt": "end_limit", "count": 12, "pct": 7.8}, {"tgt": "end_submit", "count": 7, "pct": 4.6}, {"tgt": "broke", "count": 6, "pct": 3.9}, {"tgt": "churn", "count": 4, "pct": 2.6}, {"tgt": "regressed", "count": 3, "pct": 2.0}]}, {"src": "breakthrough", "total": 362, "outs": [{"tgt": "end_submit", "count": 286, "pct": 79.0}, {"tgt": "stay", "count": 68, "pct": 18.8}, {"tgt": "regressed", "count": 6, "pct": 1.7}, {"tgt": "broke", "count": 2, "pct": 0.6}]}, {"src": "stay", "total": 302, "outs": [{"tgt": "stay", "count": 234, "pct": 77.5}, {"tgt": "end_submit", "count": 45, "pct": 14.9}, {"tgt": "end_limit", "count": 11, "pct": 3.6}, {"tgt": "broke", "count": 9, "pct": 3.0}, {"tgt": "regressed", "count": 3, "pct": 1.0}]}, {"src": "regressed", "total": 27, "outs": [{"tgt": "improved", "count": 8, "pct": 29.6}, {"tgt": "stuck", "count": 7, "pct": 25.9}, {"tgt": "churn", "count": 5, "pct": 18.5}, {"tgt": "breakthrough", "count": 4, "pct": 14.8}, {"tgt": "regressed", "count": 1, "pct": 3.7}, {"tgt": "end_limit", "count": 1, "pct": 3.7}, {"tgt": "broke", "count": 1, "pct": 3.7}]}, {"src": "broke", "total": 25, "outs": [{"tgt": "stuck", "count": 11, "pct": 44.0}, {"tgt": "improved", "count": 7, "pct": 28.0}, {"tgt": "breakthrough", "count": 6, "pct": 24.0}, {"tgt": "end_limit", "count": 1, "pct": 4.0}]}], "label_rows": [{"label": "cosmetic_refactor", "total": 108, "edges": [{"src": "stay", "tgt": "stay", "count": 79, "pct": 73.1}, {"src": "breakthrough", "tgt": "stay", "count": 14, "pct": 13.0}, {"src": "improved", "tgt": "stuck", "count": 7, "pct": 6.5}, {"src": "stuck", "tgt": "stuck", "count": 5, "pct": 4.6}, {"src": "regressed", "tgt": "stuck", "count": 1, "pct": 0.9}, {"src": "broke", "tgt": "breakthrough", "count": 1, "pct": 0.9}, {"src": "stay", "tgt": "regressed", "count": 1, "pct": 0.9}]}, {"label": "debug_failure", "total": 32, "edges": [{"src": "stuck", "tgt": "stuck", "count": 9, "pct": 28.1}, {"src": "improved", "tgt": "end_limit", "count": 5, "pct": 15.6}, {"src": "stay", "tgt": "stay", "count": 4, "pct": 12.5}, {"src": "broke", "tgt": "stuck", "count": 4, "pct": 12.5}, {"src": "improved", "tgt": "stuck", "count": 2, "pct": 6.2}, {"src": "stay", "tgt": "end_limit", "count": 2, "pct": 6.2}, {"src": "churn", "tgt": "churn", "count": 2, "pct": 6.2}, {"src": "stuck", "tgt": "broke", "count": 1, "pct": 3.1}, {"src": "regressed", "tgt": "stuck", "count": 1, "pct": 3.1}, {"src": "stuck", "tgt": "end_limit", "count": 1, "pct": 3.1}, {"src": "improved", "tgt": "broke", "count": 1, "pct": 3.1}]}, {"label": "edge_case", "total": 9, "edges": [{"src": "stay", "tgt": "stay", "count": 6, "pct": 66.7}, {"src": "breakthrough", "tgt": "stay", "count": 2, "pct": 22.2}, {"src": "regressed", "tgt": "churn", "count": 1, "pct": 11.1}]}, {"label": "execute_plan", "total": 15, "edges": [{"src": "stuck", "tgt": "improved", "count": 6, "pct": 40.0}, {"src": "stuck", "tgt": "breakthrough", "count": 4, "pct": 26.7}, {"src": "stay", "tgt": "stay", "count": 1, "pct": 6.7}, {"src": "breakthrough", "tgt": "stay", "count": 1, "pct": 6.7}, {"src": "broke", "tgt": "improved", "count": 1, "pct": 6.7}, {"src": "broke", "tgt": "stuck", "count": 1, "pct": 6.7}, {"src": "broke", "tgt": "breakthrough", "count": 1, "pct": 6.7}]}, {"label": "explore", "total": 71, "edges": [{"src": "stuck", "tgt": "stuck", "count": 17, "pct": 23.9}, {"src": "start", "tgt": "stuck", "count": 17, "pct": 23.9}, {"src": "stay", "tgt": "broke", "count": 7, "pct": 9.9}, {"src": "broke", "tgt": "stuck", "count": 5, "pct": 7.0}, {"src": "stay", "tgt": "stay", "count": 4, "pct": 5.6}, {"src": "improved", "tgt": "end_limit", "count": 4, "pct": 5.6}, {"src": "stuck", "tgt": "broke", "count": 3, "pct": 4.2}, {"src": "stuck", "tgt": "end_limit", "count": 3, "pct": 4.2}, {"src": "improved", "tgt": "broke", "count": 3, "pct": 4.2}, {"src": "improved", "tgt": "stuck", "count": 2, "pct": 2.8}, {"src": "churn", "tgt": "end_limit", "count": 2, "pct": 2.8}, {"src": "start", "tgt": "improved", "count": 1, "pct": 1.4}, {"src": "stuck", "tgt": "improved", "count": 1, "pct": 1.4}, {"src": "breakthrough", "tgt": "broke", "count": 1, "pct": 1.4}, {"src": "broke", "tgt": "end_limit", "count": 1, "pct": 1.4}]}, {"label": "fix_bug", "total": 107, "edges": [{"src": "stay", "tgt": "stay", "count": 24, "pct": 22.4}, {"src": "stuck", "tgt": "stuck", "count": 11, "pct": 10.3}, {"src": "breakthrough", "tgt": "stay", "count": 10, "pct": 9.3}, {"src": "stuck", "tgt": "improved", "count": 10, "pct": 9.3}, {"src": "improved", "tgt": "breakthrough", "count": 9, "pct": 8.4}, {"src": "stuck", "tgt": "breakthrough", "count": 8, "pct": 7.5}, {"src": "improved", "tgt": "regressed", "count": 5, "pct": 4.7}, {"src": "improved", "tgt": "improved", "count": 4, "pct": 3.7}, {"src": "broke", "tgt": "improved", "count": 3, "pct": 2.8}, {"src": "regressed", "tgt": "improved", "count": 3, "pct": 2.8}, {"src": "improved", "tgt": "stuck", "count": 2, "pct": 1.9}, {"src": "regressed", "tgt": "stuck", "count": 2, "pct": 1.9}, {"src": "regressed", "tgt": "breakthrough", "count": 2, "pct": 1.9}, {"src": "regressed", "tgt": "churn", "count": 2, "pct": 1.9}, {"src": "churn", "tgt": "churn", "count": 2, "pct": 1.9}, {"src": "churn", "tgt": "stuck", "count": 2, "pct": 1.9}, {"src": "stuck", "tgt": "churn", "count": 2, "pct": 1.9}, {"src": "improved", "tgt": "churn", "count": 1, "pct": 0.9}, {"src": "breakthrough", "tgt": "regressed", "count": 1, "pct": 0.9}, {"src": "broke", "tgt": "breakthrough", "count": 1, "pct": 0.9}, {"src": "churn", "tgt": "breakthrough", "count": 1, "pct": 0.9}, {"src": "stuck", "tgt": "regressed", "count": 1, "pct": 0.9}, {"src": "churn", "tgt": "improved", "count": 1, "pct": 0.9}]}, {"label": "implement", "total": 455, "edges": [{"src": "start", "tgt": "breakthrough", "count": 313, "pct": 68.8}, {"src": "stay", "tgt": "stay", "count": 35, "pct": 7.7}, {"src": "start", "tgt": "improved", "count": 33, "pct": 7.3}, {"src": "start", "tgt": "stuck", "count": 17, "pct": 3.7}, {"src": "breakthrough", "tgt": "stay", "count": 15, "pct": 3.3}, {"src": "stuck", "tgt": "stuck", "count": 8, "pct": 1.8}, {"src": "stuck", "tgt": "improved", "count": 6, "pct": 1.3}, {"src": "improved", "tgt": "stuck", "count": 5, "pct": 1.1}, {"src": "breakthrough", "tgt": "regressed", "count": 3, "pct": 0.7}, {"src": "stuck", "tgt": "breakthrough", "count": 2, "pct": 0.4}, {"src": "improved", "tgt": "regressed", "count": 2, "pct": 0.4}, {"src": "regressed", "tgt": "improved", "count": 2, "pct": 0.4}, {"src": "stuck", "tgt": "regressed", "count": 2, "pct": 0.4}, {"src": "churn", "tgt": "stuck", "count": 2, "pct": 0.4}, {"src": "broke", "tgt": "improved", "count": 1, "pct": 0.2}, {"src": "stuck", "tgt": "broke", "count": 1, "pct": 0.2}, {"src": "regressed", "tgt": "stuck", "count": 1, "pct": 0.2}, {"src": "broke", "tgt": "breakthrough", "count": 1, "pct": 0.2}, {"src": "regressed", "tgt": "breakthrough", "count": 1, "pct": 0.2}, {"src": "improved", "tgt": "breakthrough", "count": 1, "pct": 0.2}, {"src": "breakthrough", "tgt": "broke", "count": 1, "pct": 0.2}, {"src": "churn", "tgt": "churn", "count": 1, "pct": 0.2}, {"src": "improved", "tgt": "broke", "count": 1, "pct": 0.2}, {"src": "stuck", "tgt": "churn", "count": 1, "pct": 0.2}]}, {"label": "infra_confusion", "total": 9, "edges": [{"src": "stay", "tgt": "stay", "count": 4, "pct": 44.4}, {"src": "stay", "tgt": "end_submit", "count": 2, "pct": 22.2}, {"src": "stuck", "tgt": "breakthrough", "count": 1, "pct": 11.1}, {"src": "stay", "tgt": "broke", "count": 1, "pct": 11.1}, {"src": "improved", "tgt": "breakthrough", "count": 1, "pct": 11.1}]}, {"label": "optimize_perf", "total": 87, "edges": [{"src": "stay", "tgt": "stay", "count": 42, "pct": 48.3}, {"src": "breakthrough", "tgt": "stay", "count": 19, "pct": 21.8}, {"src": "stuck", "tgt": "breakthrough", "count": 4, "pct": 4.6}, {"src": "improved", "tgt": "stuck", "count": 3, "pct": 3.4}, {"src": "stuck", "tgt": "stuck", "count": 3, "pct": 3.4}, {"src": "improved", "tgt": "regressed", "count": 3, "pct": 3.4}, {"src": "improved", "tgt": "improved", "count": 2, "pct": 2.3}, {"src": "improved", "tgt": "breakthrough", "count": 2, "pct": 2.3}, {"src": "stay", "tgt": "regressed", "count": 2, "pct": 2.3}, {"src": "broke", "tgt": "stuck", "count": 1, "pct": 1.1}, {"src": "breakthrough", "tgt": "regressed", "count": 1, "pct": 1.1}, {"src": "regressed", "tgt": "stuck", "count": 1, "pct": 1.1}, {"src": "regressed", "tgt": "breakthrough", "count": 1, "pct": 1.1}, {"src": "regressed", "tgt": "churn", "count": 1, "pct": 1.1}, {"src": "regressed", "tgt": "improved", "count": 1, "pct": 1.1}, {"src": "regressed", "tgt": "broke", "count": 1, "pct": 1.1}]}, {"label": "revert", "total": 25, "edges": [{"src": "stay", "tgt": "stay", "count": 15, "pct": 60.0}, {"src": "stuck", "tgt": "stuck", "count": 8, "pct": 32.0}, {"src": "improved", "tgt": "stuck", "count": 1, "pct": 4.0}, {"src": "churn", "tgt": "breakthrough", "count": 1, "pct": 4.0}]}, {"label": "rewrite", "total": 54, "edges": [{"src": "stuck", "tgt": "stuck", "count": 8, "pct": 14.8}, {"src": "churn", "tgt": "churn", "count": 6, "pct": 11.1}, {"src": "stuck", "tgt": "improved", "count": 5, "pct": 9.3}, {"src": "breakthrough", "tgt": "stay", "count": 4, "pct": 7.4}, {"src": "improved", "tgt": "churn", "count": 4, "pct": 7.4}, {"src": "improved", "tgt": "regressed", "count": 4, "pct": 7.4}, {"src": "improved", "tgt": "breakthrough", "count": 3, "pct": 5.6}, {"src": "broke", "tgt": "improved", "count": 2, "pct": 3.7}, {"src": "improved", "tgt": "stuck", "count": 2, "pct": 3.7}, {"src": "churn", "tgt": "broke", "count": 2, "pct": 3.7}, {"src": "stuck", "tgt": "breakthrough", "count": 2, "pct": 3.7}, {"src": "broke", "tgt": "breakthrough", "count": 2, "pct": 3.7}, {"src": "regressed", "tgt": "improved", "count": 2, "pct": 3.7}, {"src": "stuck", "tgt": "broke", "count": 1, "pct": 1.9}, {"src": "breakthrough", "tgt": "regressed", "count": 1, "pct": 1.9}, {"src": "regressed", "tgt": "stuck", "count": 1, "pct": 1.9}, {"src": "stay", "tgt": "broke", "count": 1, "pct": 1.9}, {"src": "regressed", "tgt": "churn", "count": 1, "pct": 1.9}, {"src": "regressed", "tgt": "regressed", "count": 1, "pct": 1.9}, {"src": "churn", "tgt": "stuck", "count": 1, "pct": 1.9}, {"src": "stuck", "tgt": "churn", "count": 1, "pct": 1.9}]}, {"label": "verify", "total": 387, "edges": [{"src": "breakthrough", "tgt": "end_submit", "count": 286, "pct": 73.9}, {"src": "stay", "tgt": "end_submit", "count": 42, "pct": 10.9}, {"src": "stay", "tgt": "stay", "count": 20, "pct": 5.2}, {"src": "stay", "tgt": "end_limit", "count": 10, "pct": 2.6}, {"src": "stuck", "tgt": "end_limit", "count": 9, "pct": 2.3}, {"src": "stuck", "tgt": "end_submit", "count": 6, "pct": 1.6}, {"src": "improved", "tgt": "end_limit", "count": 4, "pct": 1.0}, {"src": "breakthrough", "tgt": "stay", "count": 3, "pct": 0.8}, {"src": "stuck", "tgt": "stuck", "count": 3, "pct": 0.8}, {"src": "churn", "tgt": "end_limit", "count": 2, "pct": 0.5}, {"src": "regressed", "tgt": "end_limit", "count": 1, "pct": 0.3}, {"src": "improved", "tgt": "end_submit", "count": 1, "pct": 0.3}]}], "n_problems": 381, "max_count": 313, "lang": "python", "model_scope": "ALL"};

	function esc(s) {
	return String(s).replace(/&/g,'&').replace(/</g,'<')
	.replace(/>/g,'>').replace(/"/g,'"');
	}
	function snapLabel(idx) {
	if (idx < 0) return idx === -1 ? 'end' : 'start';
	return 'snap_' + String(idx).padStart(2,'0');
	}

	// Action-kind palette (text / background) — must mirror Python's KIND_COLOR.
	const AKC = {
	'write': ['#166534', '#dcfce7'],
	'write_script': ['#7c3aed', '#f5f3ff'],
	'run_sample': ['#0369a1', '#e0f2fe'],
	'run_script': ['#0284c7', '#e0f2fe'],
	'test': ['#b45309', '#fffbeb'],
	'time': ['#ca8a04', '#fefce8'],
	'inspect': ['#475569', '#f1f5f9'],
	'wc': ['#64748b', '#f1f5f9'],
	'diff': ['#64748b', '#f1f5f9'],
	'submit': ['#dc2626', '#fef2f2'],
	};
	// Compact pill labels (chain display).
	const KIND_SHORT = {
	'write': 'cat > solution',
	'write_script': 'cat > helper.py',
	'run_sample': 'python3 solution.py',
	'run_script': 'python3 script.py',
	'test': './test.sh',
	'time': 'time solution.py',
	'inspect': 'cat tests/',
	'wc': 'wc',
	'diff': 'diff',
	'submit': 'echo COMPLETE_TASK',
	};
	// Gap label palette [text, background] — covers both old and new taxonomy labels
	const LABEL_COLORS = {
	// new taxonomy
	'fix_bug': ['#166534', '#dcfce7'],
	'debug_failure': ['#dc2626', '#fef2f2'],
	'implement': ['#1d4ed8', '#eff6ff'],
	'rewrite': ['#7c3aed', '#f5f3ff'],
	'explore': ['#0891b2', '#ecfeff'],
	'execute_plan': ['#0f766e', '#f0fdfa'],
	'revert': ['#6b7280', '#f9fafb'],
	'verify': ['#334155', '#f1f5f9'],
	'edge_case': ['#65a30d', '#f7fee7'],
	'optimize_perf': ['#6d28d9', '#f5f3ff'],
	'cosmetic_refactor': ['#475569', '#f1f5f9'],
	'infra_confusion': ['#ca8a04', '#fefce8'],
	// old taxonomy (v1)
	'initial_impl': ['#166534', '#dcfce7'],
	'run_tests': ['#0369a1', '#e0f2fe'],
	'fix_specific_failure':['#ea580c', '#fff7ed'],
	'rethink_algorithm': ['#d97706', '#fffbeb'],
	'complete_partial': ['#0891b2', '#ecfeff'],
	'fix_crash': ['#b91c1c', '#fef2f2'],
	'read_tests': ['#0284c7', '#f0f9ff'],
	'attempt': ['#1d4ed8', '#eff6ff'],
	};
	function renderLabelDist(labelDist, nLabeled) {
	if (!labelDist \|\| !labelDist.length) return '';
	let h = `<div class="det-sub">Gap label distribution (${nLabeled} labeled)</div>`;
	labelDist.forEach(d => {
	const [fg, bg] = LABEL_COLORS[d.label] \|\| ['#374151', '#f9fafb'];
	h += `<div class="lbl-row">`;
	h += `<div class="lbl-head">`;
	h += `<span class="lbl-pill" style="color:${fg};background:${bg}">${esc(d.label)}</span>`;
	h += `<span class="lbl-num">${d.count} · ${d.pct}%</span></div>`;
	h += `<div class="lbl-bar"><div class="lbl-fill" style="width:${d.pct}%;background:${fg}"></div></div>`;
	h += `</div>`;
	});
	return h;
	}

	function renderGapDist(dist, total, title) {
	if (!dist \|\| !Object.keys(dist).length) return '';
	const entries = Object.entries(dist).sort((a, b) => b[1] - a[1]);
	let h = `<div class="det-sub">${esc(title)} (${total} labeled)</div>`;
	entries.forEach(([lbl, pct]) => {
	const [fg, bg] = LABEL_COLORS[lbl] \|\| ['#374151', '#f9fafb'];
	h += `<div class="lbl-row">`;
	h += `<div class="lbl-head">`;
	h += `<span class="lbl-pill" style="color:${fg};background:${bg}">${esc(lbl)}</span>`;
	h += `<span class="lbl-num">${pct}%</span></div>`;
	h += `<div class="lbl-bar"><div class="lbl-fill" style="width:${pct}%;background:${fg}"></div></div>`;
	h += `</div>`;
	});
	return h;
	}

	const NLP_LABELS = {
	'm1_correctness': 'M1 correctness',
	'm1_efficiency': 'M1 efficiency',
	'm1_clarity': 'M1 clarity',
	'm2': 'M2 second-guess',
	'm3_existing': 'M3 existing tests',
	'm3_own': 'M3 own tests',
	'm4': 'M4 # comment',
	'm6': 'M6 code fence',
	'm7': 'M7 edge case',
	'm8_lang': 'M8a extra lang',
	'm8_python': 'M8b extra .py',
	'm9': 'M9 error turn',
	};
	const NLP_KEY_ORDER = ['m1_correctness','m1_efficiency','m1_clarity','m2',
	'm3_existing','m3_own','m4','m6','m7','m8_lang','m8_python','m9'];

	function renderNlpDist(rates, total, title) {
	if (!rates \|\| !Object.keys(rates).length) return '';
	const entries = NLP_KEY_ORDER
	.filter(k => (rates[k] \|\| 0) > 0)
	.map(k => [k, rates[k]]);
	if (!entries.length) return '';
	let h = `<div class="det-sub">${esc(title)} (${total} edges)</div>`;
	entries.forEach(([k, pct]) => {
	const lbl = NLP_LABELS[k] \|\| k;
	h += `<div class="nlp-row">`;
	h += `<span class="nlp-lbl">${esc(lbl)}</span>`;
	h += `<span class="nlp-bar"><span class="nlp-fill" style="width:${pct}%"></span></span>`;
	h += `<span class="nlp-num">${pct}%</span></div>`;
	});
	return h;
	}

	function renderNlpRates(rates) {
	if (!rates \|\| !Object.keys(rates).length) return '';
	const chips = NLP_KEY_ORDER
	.filter(k => (rates[k] \|\| 0) > 0)
	.map(k => `<span class="nlp-chip">${esc(NLP_LABELS[k] \|\| k)}: ${rates[k]}%</span>`)
	.join('');
	if (!chips) return '';
	return `<div class="det-sub">NLP signals (% of traversals)</div><div style="margin-bottom:6px">${chips}</div>`;
	}

	function renderChain(chain) {
	if (!chain \|\| !chain.length)
	return '<span class="kind-pill" style="color:#94a3b8;background:#f1f5f9">(no useful command)</span>';
	return chain.map((k, i) => {
	const [fg, bg] = AKC[k] \|\| ['#374151', '#f9fafb'];
	const lbl = KIND_SHORT[k] \|\| k;
	const arrow = i > 0 ? '<span class="kind-arrow">→</span>' : '';
	return arrow + `<span class="kind-pill" style="color:${fg};background:${bg}">${esc(lbl)}</span>`;
	}).join('');
	}

	function showEdgeDetail(panel, ed) {
	const sc = (M.nodes.find(n => n.id === ed.source) \|\| {}).color \|\| ed.color;
	const tc = (M.nodes.find(n => n.id === ed.target) \|\| {}).color \|\| '#9e9e9e';
	let h = `<div class="det-title">`;
	h += `<span class="state-chip" style="background:${sc}">${esc(ed.source)}</span>`;
	h += `<span style="color:#94a3b8;font-weight:400">→</span>`;
	h += `<span class="state-chip" style="background:${tc}">${esc(ed.target)}</span></div>`;
	h += `<div class="det-row"><span class="det-lbl">Count</span><span class="det-val">${ed.count}</span></div>`;
	h += `<div class="det-row"><span class="det-lbl">P(${esc(ed.target)} \| ${esc(ed.source)})</span>`;
	h += `<span class="det-val">${ed.out_pct}%</span></div>`;

	h += `<div class="det-sub">Command-chain breakdown (${ed.n_unique_chains} distinct chain${ed.n_unique_chains === 1 ? '' : 's'})</div>`;
	(ed.breakdown \|\| []).forEach(b => {
	h += `<div class="bd-row">`;
	h += `<div class="bd-head"><span class="bd-lbl">${renderChain(b.chain)}</span>`;
	h += `<span class="bd-num">${b.count} · ${b.pct}%</span></div>`;
	h += `<div class="bd-bar"><div class="bd-fill" style="width:${b.pct}%;background:#94a3b8"></div></div>`;
	(b.examples \|\| []).forEach(ex => {
	const fr = ex.snap_from < 0 ? 'start' : snapLabel(ex.snap_from);
	const to = ex.snap_to < 0
	? (ed.target === 'end_limit' ? 'limit' : 'submit')
	: snapLabel(ex.snap_to);
	h += `<div class="bd-ex">${esc(ex.problem)} ${fr}→${to}</div>`;
	});
	h += `</div>`;
	});
	h += renderLabelDist(ed.label_dist, ed.n_labeled);
	h += renderNlpRates(ed.nlp_rates);
	panel.innerHTML = h;
	}

	function showNodeDetail(panel, n) {
	let h = `<div class="det-title"><span class="state-chip" style="background:${n.color}">${n.state_type}</span></div>`;
	h += `<div class="det-def">${esc(n.definition)}</div>`;
	h += `<div class="det-row"><span class="det-lbl">Visits</span><span class="det-val">${n.visits}</span></div>`;
	h += `<div class="det-row"><span class="det-lbl">Outgoing</span><span class="det-val">${n.out_total}</span></div>`;

	const outs = M.edges.filter(e => e.source === n.id)
	.sort((a,b) => b.count - a.count);
	if (outs.length) {
	h += `<div class="det-sub">Outgoing distribution (sums to 100%)</div>`;
	outs.forEach(e => {
	const tcol = (M.nodes.find(nd => nd.id === e.target) \|\| {}).color \|\| e.color;
	h += `<div class="out-row">`;
	h += `<span class="out-tgt" style="color:${tcol}">${esc(e.target)}</span>`;
	h += `<span class="out-bar"><span class="out-fill" style="width:${e.out_pct}%;background:${tcol}"></span></span>`;
	h += `<span class="out-num">${e.out_pct}% (${e.count})</span></div>`;
	});
	} else {
	h += `<div class="det-def">Terminal state — no outgoing transitions.</div>`;
	}

	const ins = M.edges.filter(e => e.target === n.id)
	.sort((a,b) => b.count - a.count);
	if (ins.length) {
	const inTotal = ins.reduce((s, e) => s + e.count, 0);
	h += `<div class="det-sub">Incoming distribution (sums to 100%)</div>`;
	ins.forEach(e => {
	const scol = (M.nodes.find(nd => nd.id === e.source) \|\| {}).color \|\| e.color;
	const pct = inTotal > 0 ? Math.round(1000 * e.count / inTotal) / 10 : 0;
	h += `<div class="out-row">`;
	h += `<span class="out-tgt" style="color:${scol}">${esc(e.source)}</span>`;
	h += `<span class="out-bar"><span class="out-fill" style="width:${pct}%;background:${scol}"></span></span>`;
	h += `<span class="out-num">${pct}% (${e.count})</span></div>`;
	});
	}
	h += renderGapDist(n.gap_out, n.gap_out_total, 'Outgoing gap labels');
	h += renderGapDist(n.gap_in, n.gap_in_total, 'Incoming gap labels');
	h += renderNlpDist(n.nlp_out, n.nlp_out_total, 'Outgoing NLP signals');
	h += renderNlpDist(n.nlp_in, n.nlp_in_total, 'Incoming NLP signals');
	panel.innerHTML = h;
	}

	function makeBgImg(count, w, h, fs) {
	if (!count) return '';
	const svg = '<svg xmlns="http://www.w3.org/2000/svg" width="' + w + '" height="' + h + '"><text x="' + (w/2) + '" y="' + Math.round(h*0.8) + '" text-anchor="middle" font-family="ui-monospace,monospace" font-size="' + fs + '" fill="white" font-weight="700">' + count + '</text></svg>';
	return 'data:image/svg+xml;charset=utf-8,' + encodeURIComponent(svg);
	}

	const cyNodes = M.nodes.map(n => {
	const sp = n.id === 'start' \|\| n.id === 'end_submit' \|\| n.id === 'end_limit';
	const nw = sp ? 54 : 72, nh = sp ? 54 : 72;
	return {
	data: { id: n.id, label: n.label, color: n.color, n,
	bgImg: makeBgImg(n.visits, nw, nh, 9) },
	position: { x: n.x, y: n.y },
	};
	});

	const cyEdges = M.edges.map(ed => ({
	data: {
	id: ed.id, source: ed.source, target: ed.target,
	label: ed.label, color: ed.color,
	isSelf: ed.source === ed.target ? 'yes' : 'no',
	// width scales with outgoing probability (thicker = more likely)
	// edge thickness scales with ABSOLUTE count (normalised against the
	// graph's max count so the thickest edge fills the visible range).
	width: 1.0 + (ed.count / Math.max(1, M.max_count)) * 9.0,
	ed,
	}
	}));

	const cy = cytoscape({
	container: document.getElementById('cy'),
	elements: { nodes: cyNodes, edges: cyEdges },
	layout: { name: 'preset' },
	style: [
	{
	selector: 'node',
	style: {
	'background-color': 'data(color)', 'label': 'data(label)', 'color': 'white',
	'font-size': '11px', 'font-weight': '700',
	'text-valign': 'center', 'text-halign': 'center',
	'width': 72, 'height': 72, 'border-width': 3, 'border-color': 'white',
	'background-image': 'data(bgImg)',
	'background-fit': 'contain',
	}
	},
	{
	selector: 'node[id = "start"], node[id = "end_submit"], node[id = "end_limit"]',
	style: { 'shape': 'round-rectangle', 'width': 54, 'height': 54,
	'background-color': '#94a3b8', 'border-color': '#64748b' }
	},
	{
	selector: 'node[id = "start"]',
	style: { 'font-size': '16px' }
	},
	{
	selector: 'node:selected',
	style: { 'border-color': '#0f172a', 'border-width': 4 }
	},
	{
	selector: 'edge',
	style: {
	'width': 'data(width)',
	'line-color': 'data(color)',
	'target-arrow-color': 'data(color)',
	'target-arrow-shape': 'triangle',
	'arrow-scale': 1.0,
	'curve-style': 'bezier',
	'target-label': 'data(label)',
	'target-text-offset': 24,
	'font-size': '9px',
	'font-weight': '700',
	'font-family': 'ui-monospace, monospace',
	'color': '#1e293b',
	'text-background-opacity': 0.92,
	'text-background-color': 'white',
	'text-background-padding': '2px',
	'opacity': 0.78,
	}
	},
	{
	selector: 'edge[isSelf = "yes"]',
	style: { 'curve-style': 'loop', 'loop-direction': '0deg',
	'loop-sweep': '70deg' }
	},
	{
	selector: 'edge:selected',
	style: { 'opacity': 1, 'font-size': '11px', 'z-index': 999, 'line-color': '#0f172a',
	'target-arrow-color': '#0f172a' }
	},
	],
	userZoomingEnabled: true,
	userPanningEnabled: true,
	boxSelectionEnabled: false,
	});
	cy.fit(cy.elements(), 40);

	const panel = document.getElementById('detail');
	cy.on('tap', 'node', evt => showNodeDetail(panel, evt.target.data('n')));
	cy.on('tap', 'edge', evt => showEdgeDetail(panel, evt.target.data('ed')));
	cy.on('tap', evt => {
	if (evt.target === cy)
	panel.innerHTML = '<div class="detail-placeholder">Click a node or edge for details</div>';
	});
	cy.on('mouseover', 'edge', evt => evt.target.style({ opacity: 1 }));
	cy.on('mouseout', 'edge', evt => { if (!evt.target.selected()) evt.target.style({ opacity: 0.78 }); });
	</script>
	</body>
	</html>