contextualfailures-website/index.html at master · MIT-REALM/contextualfailures-website · GitHub

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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="utf-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1" />
  <title>Contextual Failures • CoRL 2025</title>
  <meta name="description" content="Project page for ‘COst aware Discovery of Contextual Failure using Bayesian Active Learning’ (CoRL 2025)." />
  <link rel="preconnect" href="https://fonts.googleapis.com">
  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
  <link href="https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700;800;900&display=swap" rel="stylesheet">
  <script src="https://cdn.tailwindcss.com"></script>
  <style> html { scroll-behavior: smooth } body { font-family: Inter, system-ui, -apple-system, Segoe UI, Roboto, Arial, sans-serif } code, pre { font-family: ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, "Liberation Mono", monospace } </style>
</head>
<body class="bg-neutral-50 text-neutral-900">
  <!-- Sticky nav -->
  <header class="sticky top-0 z-40 backdrop-blur bg-white/80 border-b border-neutral-200">
    <div class="max-w-6xl mx-auto px-4 sm:px-6 py-3 flex items-center justify-between">
      <a href="#top" class="font-semibold">Project • CoRL 2025</a>
      <nav class="hidden md:flex gap-5 text-sm">
        <a href="#abstract" class="hover:underline">Abstract</a>
        <!--a href="#video" class="hover:underline">Video</a-->
        <a href="#results" class="hover:underline">Results</a>
        <a href="#code" class="hover:underline">Code</a>
        <a href="#bibtex" class="hover:underline">BibTeX</a>
        <a href="#team" class="hover:underline">Team</a>
      </nav>
    </div>
  </header>

  <!-- Hero -->
  <section id="top" class="relative">
    <div class="max-w-6xl mx-auto px-4 sm:px-6">
      <div class="py-12 md:py-20">
        <div class="text-center">
          <p class="text-sm uppercase tracking-widest text-neutral-500">Conference on Robot Learning • CoRL 2025</p>
          <h1 class="mt-3 text-3xl sm:text-4xl md:text-6xl font-extrabold">Cost-aware Discovery of Contextual Failures using Bayesian Active Learning</h1>
          <p class="mt-4 text-lg text-neutral-600 max-w-3xl mx-auto">An active learning strategy for contextual failure discovery using limited data with expert in loop</p>
          <div class="mt-6 flex flex-wrap items-center justify-center gap-3">
            <a class="px-4 py-2 rounded-xl bg-black text-white" href="https://openreview.net/pdf/0c4bbc4ecee323522e632879cc6fa741268a8c13.pdf">PDF</a>
            <a class="px-4 py-2 rounded-xl bg-neutral-200" href="#code">Code</a>
          </div>
          <div class="mt-6 text-sm text-neutral-500">Anjali Parashar*, Joseph Zhang, Yinkge Li, Chuchu Fan</div>
          <img class="mt-10 rounded-2xl shadow" src="static/images/overview.png" alt="Teaser figure" />
          <p class="mt-2 text-xs text-neutral-500">Example of a contextual failure discovered using our method</p>
        </div>
      </div>
    </div>
  </section>

  <div class="max-w-6xl mx-auto px-4 sm:px-6 py-8">
    <p class="mt-2 text-neutral-700">Failure examples for Diffusion Policy trained in Sim for Push-T task on UR3E.</p>
    <div class="grid grid-cols-1 md:grid-cols-3 gap-6">
      <!-- Video 1 -->
      <figure class="rounded-2xl overflow-hidden bg-neutral-100 shadow">
        <video
          class="w-full"
          style="aspect-ratio: 12/14;"
          controls
          playsinline
          preload="metadata"
          poster="static/images/Mode_2.png">
          <source src="static/videos/CORL_Mode2_PushT.mp4"  type="video/mp4"  />
          Your browser doesn’t support HTML5 video.
        </video>
        <figcaption class="p-3 text-sm text-neutral-600">Discovered failure due to training data (Diffusion Policy: Push-T)</figcaption>
      </figure>

      <!-- Video 2 -->
      <figure class="rounded-2xl overflow-hidden bg-neutral-100 shadow">
        <video
          class="w-full"
          style="aspect-ratio: 12/14;"
          controls
          playsinline
          preload="metadata"
          poster="static/images/Mode_3.png">
          <source src="static/videos/CORL_Mode3_PushT.mp4"  type="video/mp4"  />
          Your browser doesn’t support HTML5 video.
        </video>
        <figcaption class="p-3 text-sm text-neutral-600">Discovered failure due to joint limits (Diffusion Policy: Push-T)</figcaption>
      </figure>

      <!-- Video 3 -->
      <figure class="rounded-2xl overflow-hidden bg-neutral-100 shadow">
        <video
          class="w-full"
          style="aspect-ratio:12/14;"
          controls
          playsinline
          preload="metadata"
          poster="static/images/Repair.png">
          <source src="static/videos/Fail_and_Repair_PushT.mp4"  type="video/mp4"  />
          Your browser doesn’t support HTML5 video.
        </video>
        <figcaption class="p-3 text-sm text-neutral-600">Failure used for repair by supplying better demonstrations</figcaption>
      </figure>
    </div>
  </div>

  <!-- Results -->
  <section id="results" class="border-t border-neutral-200 bg-white">
    <div class="max-w-6xl mx-auto px-4 sm:px-6 py-12">
      <!--h2 class="text-2xl md:text-3xl font-bold">Results</h2-->
      <p class="mt-2 text-neutral-700">Failure examples for YOLO self-driving object detection in CARLA.</p>
      <div class="mt-6 grid sm:grid-cols-2 lg:grid-cols-3 gap-6">
        <!-- Example card -->
        <figure class="bg-neutral-100 rounded-2xl p-3 shadow">
          <img class="rounded-xl" src="static/images/light.png" alt="Result 1" />
          <figcaption class="mt-2 text-sm text-neutral-600">Failure due to bad lighting</figcaption>
        </figure>
        <figure class="bg-neutral-100 rounded-2xl p-3 shadow">
          <img class="rounded-xl" src="static/images/distance.png" alt="Result 2" />
          <figcaption class="mt-2 text-sm text-neutral-600">Failure due to large distance</figcaption>
        </figure>
        <figure class="bg-neutral-100 rounded-2xl p-3 shadow">
          <img class="rounded-xl" src="static/images/distance_and_light.png" alt="Result 3" />
          <figcaption class="mt-2 text-sm text-neutral-600">Failure due to large distance and bad lighting</figcaption>
        </figure>
      </div>


    </div>
  </section>
  <!-- Put video.mp4 (H.264), video.webm (VP9/AV1), and captions.vtt in the same folder as this HTML -->
  <!--div style="max-width: 800px; margin: 1rem auto;">
    <video
      controls
      playsinline
      preload="metadata"
      style="width: 100%; height: auto; border-radius: 12px; display: block;"
      poster="thumbnail.jpg"
    >

      <source src="video.webm" type="video/webm" />
      <source src="video.mp4"  type="video/mp4" />

      <track
        src="captions.vtt"
        kind="subtitles"
        srclang="en"
        label="English"
        default
      />

      Your browser doesn’t support HTML5 video.
    </video>
  </div-->

  <!-- Video / Demo -->
  <!--section id="video" class="border-t border-neutral-200">
    <div class="max-w-5xl mx-auto px-4 sm:px-6 py-12">
      <h2 class="text-2xl md:text-3xl font-bold">Overview Video</h2>
      <p class="mt-2 text-neutral-700">Embed your 3–6 minute project video or live demo below.</p>
      <div class="mt-6 aspect-video rounded-2xl overflow-hidden shadow bg-black"-->
        <!-- Replace YOUTUBE_ID -->
        <!--iframe class="w-full h-full" src="https://www.youtube.com/embed/YOUTUBE_ID" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen></iframe>
      </div>
    </div>
  </section-->
  <!-- Abstract -->
  <section id="abstract" class="border-t border-neutral-200 bg-white">
    <div class="max-w-4xl mx-auto px-4 sm:px-6 py-12">
      <h2 class="text-2xl md:text-3xl font-bold">Abstract</h2>
      <p class="mt-4 text-neutral-700">Ensuring the robustness of robotic systems is crucial for their de- ployment in safety-critical domains. Failure discovery, or falsification, is a widely used approach for evaluating robustness, with recent advancements fo- cusing on improving sample efficiency and generalization through probabilistic sampling techniques and learning-theoretic approaches. However, existing methods rely on explicitly defined analytical cost functions to characterize failures, often overlooking the underlying causes and diversity of discovered failure scenarios. In this work, we propose a novel failure discovery framework that integrates contextual reasoning in the falsification process, specifically tailored for high evaluation-cost applications. Our method incorporates expert-in-the-loop feedback to construct a probabilistic surrogate model of failures using Bayesian inference. This model is iteratively refined and leveraged to guide an active learning strategy that prioritizes the discovery of diverse failure cases. We empirically validate our approach across a range of tasks for high-cost contextual falsification in robotic manipulation and autonomous driving.</p>
      <!--ul class="mt-4 list-disc pl-6 text-neutral-700">
        <li>Key takeaway 1.</li>
        <li>Key takeaway 2.</li>
        <li>Key takeaway 3.</li>
      </ul-->
    </div>
  </section>


  <!-- Code & Data -->
  <section id="code" class="border-t border-neutral-200">
    <div class="max-w-4xl mx-auto px-4 sm:px-6 py-12">
      <h2 class="text-2xl md:text-3xl font-bold">Code</h2>
      <p class="mt-2 text-neutral-700">Reproduce our CARLA and Push-T (sim) results</p>
      <div class="mt-4 grid md:grid-cols-2 gap-4">
        <div class="bg-neutral-100 rounded-2xl p-4">
          <h3 class="font-semibold">Quickstart</h3>
          <pre class="mt-2 text-sm overflow-x-auto"><code>git clone https://github.com/MIT-REALM/contextual_failure.git
cd contextual_failure
        </div>
  </section>

  <!-- BibTeX -->
  <section id="bibtex" class="border-t border-neutral-200 bg-white">
    <div class="max-w-4xl mx-auto px-4 sm:px-6 py-12">
      <h2 class="text-2xl md:text-3xl font-bold">BibTeX</h2>
      <pre class="mt-4 bg-neutral-100 p-4 rounded-2xl overflow-x-auto text-sm"><code>@inproceedings{parashar2025paper,
  title={Paper Title},
  author={Parashar, Anjali and Coauthors},
  booktitle={Proceedings of the Conference on Robot Learning (CoRL)},
  year={2025}
}</code></pre>
    </div>
  </section>

  <!-- Team -->
  <section id="team" class="border-t border-neutral-200">
    <div class="max-w-6xl mx-auto px-4 sm:px-6 py-12">
      <h2 class="text-2xl md:text-3xl font-bold">Authors & Affiliations</h2>
      <div class="mt-6 grid sm:grid-cols-2 lg:grid-cols-3 gap-6">
        <div class="bg-white rounded-2xl p-4 shadow">
          <div class="font-semibold">Anjali Parashar</div>
          <div class="text-sm text-neutral-600">MIT REALM Lab</div>
          <div class="mt-2 text-sm"><a class="underline" href="https://aeroastro.mit.edu/realm/team/anjali-parashar/">Homepage</a> • <a class="underline" href="https://scholar.google.com/citations?user=XMr6ploAAAAJ&hl=en&oi=ao" href="#">Scholar</a> • <a class="underline" href="anjalip@mit.edu">Email</a></div>
        </div>
        <!-- Duplicate blocks for co-authors -->
      </div>
    </div>
  </section>


  <!-- Footer -->
  <footer class="border-t border-neutral-200">
    <div class="max-w-6xl mx-auto px-4 sm:px-6 py-8 text-sm text-neutral-600 flex flex-col sm:flex-row gap-2 sm:items-center sm:justify-between">
      <div>© 2025 Authors. Please cite if you use this work.</div>
      <div class="flex gap-4">
        <a class="underline" href="https://openreview.net/pdf/0c4bbc4ecee323522e632879cc6fa741268a8c13.pdf">Paper</a>
        <a class="underline" href="https://github.com/MIT-REALM/contextual_failure">Code</a>
      </div>
    </div>
  </footer>
</body>
</html>