-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtask.html
More file actions
136 lines (126 loc) · 9.82 KB
/
Copy pathtask.html
File metadata and controls
136 lines (126 loc) · 9.82 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
<!doctype html>
<html lang="en">
<head>
<meta charset="UTF-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<meta name="description" content="Build a runnable Agent that diagnoses and repairs real cross-architecture package build failures under executable verification." />
<title>Challenge | Build-Bench</title>
<link rel="canonical" href="https://aiops-lab-nku.github.io/BuildBench/task.html" />
<link rel="icon" type="image/svg+xml" href="assets/logo.svg" />
<link rel="alternate icon" type="image/png" href="assets/favicon.png" />
<link rel="stylesheet" href="styles.css?v=20260810-7" />
<script src="assets/vendor/lucide.min.js?v=20260810-7" defer></script>
<script src="i18n/task.js?v=20260810-2" defer></script>
<script src="i18n.js?v=20260810-7" defer></script>
<script src="auth-client.js?v=20260810-7" defer></script>
<script src="app.js?v=20260810-7" defer></script>
</head>
<body class="task-page docs-page" data-page="task">
<a class="skip-link" href="#main-content">Skip to main content</a>
<header class="site-header">
<div class="header-inner">
<a class="brand" href="index.html" aria-label="Build-Bench Challenge home"><span class="brand-mark" aria-hidden="true"><img src="assets/logo.svg" alt="" /></span><span>Build-Bench</span></a>
<button class="icon-button menu-button" type="button" aria-label="Open navigation" aria-expanded="false" aria-controls="site-nav" title="Open navigation" data-menu-button><i data-lucide="menu" aria-hidden="true"></i></button>
<nav class="site-nav" id="site-nav" aria-label="Primary navigation" data-nav>
<a href="index.html">Overview</a>
<a class="active" href="task.html" aria-current="page">Challenge</a>
<a href="submission.html">Participate</a>
<a href="evaluation.html">Evaluation</a>
<a class="nav-emphasis" href="leaderboard.html"><i data-lucide="trophy" aria-hidden="true"></i>Leaderboard</a>
<a href="rules.html">Rules</a>
<div class="nav-more" data-nav-more>
<button class="nav-more-toggle" type="button" aria-expanded="false" aria-controls="nav-more-menu" data-nav-more-toggle>
<span>More</span><span class="nav-more-caret" aria-hidden="true">▾</span>
</button>
<div class="nav-more-menu" id="nav-more-menu" data-nav-more-menu>
<a href="timeline.html"><span>Timeline</span><small>Competition dates and milestones</small></a>
<a href="faq.html"><span>FAQ</span><small>Common questions and participant support</small></a>
</div>
</div>
</nav>
</div>
</header>
<main id="main-content">
<section class="page-masthead task-masthead" aria-labelledby="challenge-page-title">
<div class="content-wrap">
<div class="task-masthead-copy">
<span class="page-kicker">Agent competition</span>
<h1 id="challenge-page-title">Challenge</h1>
<p class="challenge-headline">Build an Agent that repairs real cross-architecture build failures</p>
<p>Build-Bench challenges teams to develop a runnable software-repair Agent for real packages that build successfully on one instruction-set architecture but fail when rebuilt on another.</p>
<p>For each Case, the Agent investigates a failed build inside a prepared package workspace, modifies the permitted package files, and produces a repair that can be independently verified by the competition platform.</p>
<p class="challenge-submission-principle">You submit a runnable Agent — not precomputed Case-by-Case patches.</p>
</div>
</div>
</section>
<div class="content-wrap task-doc-layout">
<aside class="page-rail task-toc" aria-label="On this page">
<strong>On this page</strong>
<nav>
<a href="#why-this-challenge">Why this challenge?</a>
<a href="#your-mission">Your mission</a>
<a href="#agent-inputs">What your Agent works with</a>
<a href="#agent-actions">What your Agent can do</a>
<a href="#repair-lifecycle">The repair lifecycle</a>
<a href="#solved">What counts as solved?</a>
<a href="#start-building">Start building</a>
</nav>
</aside>
<article class="page-document task-document">
<section class="docs-section" id="why-this-challenge">
<h2>Why this challenge?</h2>
<p>Modern software increasingly needs to run across heterogeneous architectures, toolchains, dependencies, and packaging environments. A package that works on one architecture may fail on another because of architecture-specific code, build configuration, dependencies, compiler behavior, or packaging rules.</p>
<p class="challenge-question">Can an autonomous Agent diagnose a real build failure and repair the package end to end?</p>
<p>The challenge focuses on repairs that work in a real build environment, rather than changes that merely look similar to a reference solution.</p>
</section>
<section class="docs-section" id="your-mission">
<h2>Your mission</h2>
<p>Each Case starts from a real software package that builds successfully on a source architecture but fails on a target architecture.</p>
<p>Your Agent receives a prepared package workspace and the initial failure context. It must investigate the cause of the failure and modify the permitted package files so that the package can be rebuilt successfully for the target architecture.</p>
<p>During official evaluation, the Agent must complete this task autonomously under the competition runtime and rules.</p>
</section>
<section class="docs-section" id="agent-inputs">
<h2>What your Agent works with</h2>
<p>At the beginning of a Case, the Agent is given the package workspace, source and target architecture information, the initial failed-build evidence, and the packaging and build context included with that Case.</p>
<p>The workspace may contain source code, packaging files, build scripts, existing package-side patches, metadata, and other files relevant to diagnosing the failure.</p>
<p class="challenge-section-link"><a href="submission.html#runtime-interface">See the detailed, versioned runtime specification <span aria-hidden="true">→</span></a></p>
</section>
<section class="docs-section" id="agent-actions">
<h2>What your Agent can do</h2>
<p>The Agent may inspect the available package and build context, reason about the failure, and modify files within the permitted package paths using the tools available in the official runtime.</p>
<p>Runtime limits, network policy, resource limits, allowed modification scope, and other execution constraints are defined separately in the competition documentation.</p>
<p class="challenge-inline-links"><a href="submission.html">Read the Submission Guide <span aria-hidden="true">→</span></a><a href="rules.html">Read the Rules <span aria-hidden="true">→</span></a></p>
</section>
<section class="docs-section" id="repair-lifecycle">
<h2>The repair lifecycle</h2>
<ol class="challenge-lifecycle" aria-label="Repair lifecycle">
<li>Case workspace</li>
<li>Agent diagnosis & repair</li>
<li>Final workspace</li>
<li>Clean target rebuild</li>
</ol>
<p>The organizer runs each qualified Agent in a controlled Case workspace. After the run, the platform captures the Agent's final changes and derives the repair used for verification.</p>
<p>The repair is then applied to a clean copy of the Case and evaluated in the official target-architecture build environment.</p>
</section>
<section class="docs-section" id="solved">
<h2>What counts as solved?</h2>
<p>A Case is solved only when the Agent's final changes can be applied to a clean Case and the package builds successfully in the official target-architecture environment.</p>
<p>Repairs are judged by verified build results, not by similarity to a reference patch.</p>
<p class="challenge-section-link"><a href="evaluation.html">Read the Evaluation Protocol <span aria-hidden="true">→</span></a></p>
</section>
<section class="docs-section challenge-start" id="start-building">
<h2>Start building</h2>
<p class="challenge-ready">Ready to build your Agent?</p>
<p>Use the Starter Kit and example Cases to understand the workspace, test your Agent locally, and prepare a qualified submission.</p>
<nav class="challenge-start-links" aria-label="Start building links">
<a href="data-downloads.html">Get the Starter Kit <span aria-hidden="true">→</span></a>
<a href="submission.html">Read the Submission Guide <span aria-hidden="true">→</span></a>
<a href="evaluation.html">Read the Evaluation Protocol <span aria-hidden="true">→</span></a>
</nav>
</section>
</article>
</div>
</main>
<footer class="site-footer"><div class="footer-inner"><div class="footer-brand"><span class="brand-mark" aria-hidden="true"><img src="assets/logo.svg" alt="" /></span><div><strong>Build-Bench Challenge</strong><small>ICSE 2027 Competition Track</small></div></div><nav class="footer-links" aria-label="Footer navigation"><a href="submission.html">Participate</a><a href="evaluation.html">Evaluation</a><a href="timeline.html">Timeline</a><a href="faq.html">FAQ</a></nav><a href="#main-content">Back to top<i data-lucide="arrow-up" aria-hidden="true"></i></a></div></footer>
</body>
</html>