rlhf/index.html at main · lasgroup/rlhf · 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
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="utf-8">
  <meta name="viewport" content="width=device-width, initial-scale=1">

  <!-- Primary Meta Tags -->
  <title>ActiveUltraFeedback & RewardUQ | ETH Zurich</title>
  <meta
    name="title"
    content="ActiveUltraFeedback & RewardUQ | ETH Zurich">
  <meta
    name="description"
    content="Project landing page for two RLHF papers on efficient preference data generation and uncertainty-aware reward modeling for large language models.">
  <meta
    name="keywords"
    content="RLHF, large language models, preference learning, active learning, reward modeling, uncertainty quantification, LLM alignment, ActiveUltraFeedback, RewardUQ">
  <meta
    name="author"
    content="Davit Melikidze, Marian Schneider, Jessica Lam, Martin Wertich, Daniel Yang, Samuel Stante, Florian Redhardt, Lena Libon, Parnian Kassraie, Ido Hakimi, Barna Pásztor, Andreas Krause">
  <meta name="robots" content="index, follow">
  <meta name="language" content="English">
  <meta name="theme-color" content="#2563eb">
  <meta name="msapplication-TileColor" content="#2563eb">
  <meta name="apple-mobile-web-app-capable" content="yes">
  <meta name="apple-mobile-web-app-status-bar-style" content="default">

  <!-- Open Graph -->
  <meta property="og:type" content="website">
  <meta property="og:site_name" content="ETH Zurich">
  <meta property="og:title" content="ActiveUltraFeedback & RewardUQ">
  <meta
    property="og:description"
    content="Project landing page for two RLHF papers on efficient preference data generation and uncertainty-aware reward modeling for large language models.">

  <!-- Optional: replace these with your real deployed URLs if available -->
  <!--
  <meta property="og:url" content="https://YOUR_GITHUB_USERNAME.github.io/YOUR_REPO/">
  <meta property="og:image" content="https://YOUR_GITHUB_USERNAME.github.io/YOUR_REPO/static/images/social_preview.png">
  <meta property="og:image:width" content="1200">
  <meta property="og:image:height" content="630">
  <meta property="og:image:alt" content="ActiveUltraFeedback & RewardUQ project page preview">
  -->

  <!-- Twitter -->
  <meta name="twitter:card" content="summary_large_image">
  <meta name="twitter:title" content="ActiveUltraFeedback & RewardUQ">
  <meta
    name="twitter:description"
    content="Project landing page for two RLHF papers on efficient preference data generation and uncertainty-aware reward modeling for large language models.">

  <!-- Optional: add these only if you actually have them -->
  <!--
  <meta name="twitter:site" content="@YOUR_LAB_HANDLE">
  <meta name="twitter:creator" content="@YOUR_HANDLE">
  <meta name="twitter:image" content="https://YOUR_GITHUB_USERNAME.github.io/YOUR_REPO/static/images/social_preview.png">
  <meta name="twitter:image:alt" content="ActiveUltraFeedback & RewardUQ project page preview">
  -->

  <!-- Preconnect -->
  <link rel="preconnect" href="https://fonts.googleapis.com">
  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
  <link rel="preconnect" href="https://cdn.jsdelivr.net">

  <!-- Favicon -->
  <link rel="icon" type="image/x-icon" href="static/images/favicon.ico">
  <link rel="apple-touch-icon" href="static/images/favicon.ico">

  <!-- CSS -->
  <link rel="stylesheet" href="static/css/bulma.min.css">
  <link rel="stylesheet" href="static/css/index.css">
  <link rel="preload" href="static/css/fontawesome.all.min.css" as="style" onload="this.onload=null;this.rel='stylesheet'">
  <link rel="preload" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css" as="style" onload="this.onload=null;this.rel='stylesheet'">
  <noscript>
    <link rel="stylesheet" href="static/css/fontawesome.all.min.css">
    <link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
  </noscript>

  <!-- Fonts -->
  <link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700;800&display=swap" rel="stylesheet">

  <!-- JS -->
  <script defer src="static/js/fontawesome.all.min.js"></script>

  <!-- Structured Data: landing page with two scholarly articles -->
  <script type="application/ld+json">
  {
    "@context": "https://schema.org",
    "@type": "CollectionPage",
    "name": "ActiveUltraFeedback & RewardUQ",
    "description": "Project landing page for two RLHF papers on efficient preference data generation and uncertainty-aware reward modeling for large language models.",
    "publisher": {
      "@type": "Organization",
      "name": "ETH Zurich"
    },
    "hasPart": [
      {
        "@type": "ScholarlyArticle",
        "name": "ActiveUltraFeedback: Efficient Preference Data Generation using Active Learning",
        "author": [
          { "@type": "Person", "name": "Davit Melikidze" },
          { "@type": "Person", "name": "Marian Schneider" },
          { "@type": "Person", "name": "Jessica Lam" },
          { "@type": "Person", "name": "Martin Wertich" },
          { "@type": "Person", "name": "Ido Hakimi" },
          { "@type": "Person", "name": "Barna Pásztor" },
          { "@type": "Person", "name": "Andreas Krause" }
        ],
        "sameAs": [
          "https://arxiv.org/abs/2603.09692",
          "https://github.com/lasgroup/ActiveUltraFeedback"
        ]
      },
      {
        "@type": "ScholarlyArticle",
        "name": "RewardUQ: A Unified Framework for Uncertainty-Aware Reward Models",
        "author": [
          { "@type": "Person", "name": "Daniel Yang" },
          { "@type": "Person", "name": "Samuel Stante" },
          { "@type": "Person", "name": "Florian Redhardt" },
          { "@type": "Person", "name": "Lena Libon" },
          { "@type": "Person", "name": "Parnian Kassraie" },
          { "@type": "Person", "name": "Ido Hakimi" },
          { "@type": "Person", "name": "Barna Pásztor" },
          { "@type": "Person", "name": "Andreas Krause" }
        ],
        "sameAs": [
          "https://arxiv.org/abs/2602.24040",
          "https://github.com/lasgroup/rewarduq"
        ]
      }
    ]
  }
  </script>

  <style>
    body {
      font-family: "Inter", sans-serif;
    }

    .hero.is-index {
      padding-top: 3rem;
      padding-bottom: 1.5rem;
    }

    .publication-subtitle {
      max-width: 850px;
      margin: 1rem auto 0;
      color: #4b5563;
      font-size: 1.05rem;
      line-height: 1.7;
    }

    .papers-section {
      padding-top: 1rem;
      padding-bottom: 4rem;
    }

    .paper-card {
      height: 100%;
      border-radius: 20px;
      overflow: hidden;
      border: 1px solid #e5e7eb;
      box-shadow: 0 10px 30px rgba(15, 23, 42, 0.06);
      background: #fff;
      display: flex;
      flex-direction: column;
    }

    .paper-card-image-link {
      display: block;
      border-bottom: 1px solid #e5e7eb;
      background: #fff;
    }

    .paper-card-image {
      display: block;
      width: 100%;
      aspect-ratio: 16 / 7.5;
      object-fit: contain;
      background: #fff;
      padding: 0.5rem;
    }

    .paper-card-content {
      display: flex;
      flex-direction: column;
      gap: 0.9rem;
      padding: 1.5rem;
      height: 100%;
    }

    .paper-tag {
      width: fit-content;
      border-radius: 999px;
      font-weight: 600;
      font-size: 0.85rem;
      padding: 0.35rem 0.7rem;
      background: #eff6ff;
      color: #1d4ed8;
    }

    .paper-title-link {
      color: inherit;
      text-decoration: none;
    }

    .paper-title-link:hover {
      color: #2563eb;
    }

    .paper-summary {
      color: #4b5563;
      line-height: 1.7;
      flex-grow: 1;
    }

    .paper-actions {
      display: flex;
      flex-wrap: wrap;
      gap: 0.75rem;
      margin-top: auto;
      padding-top: 0.25rem;
    }

    .copy-bibtex-btn-inline {
      border: 1px solid #dbdbdb;
      background: #fff;
      border-radius: 999px;
      padding: 0.5rem 0.95rem;
      cursor: pointer;
      font: inherit;
    }

    .copy-bibtex-btn-inline:hover {
      border-color: #b5b5b5;
    }

    .footer {
      padding-top: 2rem;
      padding-bottom: 2rem;
    }

    .institution-strip {
  padding-top: 0.75rem;
  padding-bottom: 2.5rem;
}

.institution-logos {
  display: grid;
  grid-template-columns: repeat(3, minmax(0, 1fr));
  align-items: center;
  column-gap: 2rem;
  max-width: 900px;
  margin: 0 auto;
}

.logo-box {
  height: 70px;
  display: flex;
  align-items: center;
  justify-content: center;
}

.logo-box img {
  display: block;
  max-height: 54px;
  max-width: 220px;
  width: auto;
  object-fit: contain;
}

@media screen and (max-width: 768px) {
  .institution-logos {
    column-gap: 1rem;
    max-width: 100%;
  }

  .logo-box {
    height: 56px;
  }

  .logo-box img {
    max-height: 40px;
    max-width: 100%;
  }
}
  </style>
</head>

<body>
  <main id="main-content">
    <section class="hero is-index">
      <div class="hero-body">
        <div class="container landing-wide">
          <div class="columns is-centered">
            <div class="column has-text-centered">
              <h1 class="title is-1 publication-title">ActiveUltraFeedback &amp; RewardUQ</h1>
              <p class="publication-subtitle">
                Two recent papers on RLHF for large language models:
                <strong>ActiveUltraFeedback</strong> on efficient preference data generation and
                <strong>RewardUQ</strong> on uncertainty-aware reward modeling.
                Use the cards below to navigate to the full project pages.
              </p>
            </div>
          </div>
        </div>
      </div>
    </section>

    <section class="section papers-section">
      <div class="container landing-wide">
        <div class="columns is-variable is-5">

          <!-- ActiveUltraFeedback -->
          <div class="column is-6">
            <article class="paper-card">
              <a class="paper-card-image-link" href="./ActiveUltraFeedback.html">
                <img
                  class="paper-card-image"
                  src="./static/figures/ActiveUltraFeedback/overview.png"
                  alt="Preview figure for ActiveUltraFeedback">
              </a>

              <div class="paper-card-content">
                <span class="paper-tag">Active Learning for Preference Data Generation</span>

                <h2 class="title is-4 mb-1">
                  <a class="paper-title-link" href="./ActiveUltraFeedback.html">
                    ActiveUltraFeedback: Efficient Preference Data Generation using Active Learning
                  </a>
                </h2>

                <p class="paper-summary">
                  ActiveUltraFeedback is a modular pipeline for collecting high-quality preference data more efficiently.
                  By using uncertainty-aware response selection, it identifies informative response pairs for annotation
                  and can achieve strong downstream performance with substantially fewer labels than static baselines.
                </p>

                <div class="paper-actions">
                  <a class="button-primary" href="./ActiveUltraFeedback.html">
                    <span>Project Page</span>
                  </a>

                  <a class="button-secondary" href="https://arxiv.org/abs/2603.09692" target="_blank" rel="noopener">
                    <span>Paper (arXiv)</span>
                  </a>

                  <a class="button-secondary" href="https://github.com/lasgroup/ActiveUltraFeedback" target="_blank" rel="noopener">
                    <span>GitHub Code</span>
                  </a>

                  <a
                    class="button-tertiary"
                    href="#"
                    onclick="copyBibtex(this, 'activeuf'); return false;">
                    BibTeX
                  </a>
                </div>
              </div>
            </article>
          </div>

          <!-- RewardUQ -->
          <div class="column is-6">
            <article class="paper-card">
              <a class="paper-card-image-link" href="./RewardUQ.html">
                <img
                  class="paper-card-image"
                  src="./static/figures/RewardUQ/overview_uq.png"
                  alt="Preview figure for RewardUQ">
              </a>

              <div class="paper-card-content">
                <span class="paper-tag">Uncertainty Quantification for Reward Models</span>

                <h2 class="title is-4 mb-1">
                  <a class="paper-title-link" href="./RewardUQ.html">
                    RewardUQ: A Unified Framework for Uncertainty-Aware Reward Models
                  </a>
                </h2>

                <p class="paper-summary">
                  RewardUQ provides a unified framework for evaluating uncertainty quantification in reward models.
                  It compares common methods in terms of accuracy and calibration, helping clarify which design choices
                  matter most for reliable reward modeling in LLM alignment.
                </p>

                <div class="paper-actions">
                  <a class="button-primary" href="./RewardUQ.html">
                    <span>Project Page</span>
                  </a>

                  <a class="button-secondary" href="https://arxiv.org/abs/2602.24040" target="_blank" rel="noopener">
                    <span>Paper (arXiv)</span>
                  </a>

                  <a class="button-secondary" href="https://github.com/lasgroup/rewarduq" target="_blank" rel="noopener">
                    <span>GitHub Code</span>
                  </a>

                  <a
                    class="button-tertiary"
                    href="#"
                    onclick="copyBibtex(this, 'rewarduq'); return false;">
                    BibTeX
                  </a>
                </div>
              </div>
            </article>
          </div>

        </div>
      </div>
    </section>
  </main>

  <section class="section institution-strip">
    <div class="container">
      <div class="institution-logos">
        <div class="logo-box">
          <img src="./static/images/ethz_logo.png" alt="ETH Zurich logo">
        </div>
        <div class="logo-box">
          <img src="./static/images/las_logo.png" alt="Learning & Adaptive Systems Group logo">
        </div>
        <div class="logo-box">
          <img src="./static/images/eth_ai_center_logo.png" alt="ETH AI Center logo">
        </div>
      </div>
    </div>
  </section>

  <footer class="footer">
    <div class="container">
      <div class="columns is-centered">
        <div class="column is-8">
          <div class="content has-text-centered">
            <p>
              Website and project are part of the Learning & Adaptive Systems Group at ETH Zurich and the ETH AI Center. <br>
              This page was built using the <a href="https://github.com/eliahuhorwitz/Academic-project-page-template" target="_blank">Academic Project Page Template</a>.
            </p>
          </div>
        </div>
      </div>
    </div>
  </footer>

  <script>
    const bibtexEntries = {
      activeuf: `@misc{melikidze2026activeultrafeedbackefficientpreferencedata,
  title={ActiveUltraFeedback: Efficient Preference Data Generation using Active Learning},
  author={Davit Melikidze and Marian Schneider and Jessica Lam and Martin Wertich and Ido Hakimi and Barna Pásztor and Andreas Krause},
  year={2026},
  eprint={2603.09692},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2603.09692},
}`,
      rewarduq: `@misc{yang2026rewarduqunifiedframeworkuncertaintyaware,
  title={RewardUQ: A Unified Framework for Uncertainty-Aware Reward Models},
  author={Daniel Yang and Samuel Stante and Florian Redhardt and Lena Libon and Parnian Kassraie and Ido Hakimi and Barna Pásztor and Andreas Krause},
  year={2026},
  eprint={2602.24040},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2602.24040},
}`
    };

    async function copyBibtex(button, key) {
      const text = bibtexEntries[key];
      const original = button.textContent;

      try {
        if (navigator.clipboard && window.isSecureContext) {
          await navigator.clipboard.writeText(text);
        } else {
          const textArea = document.createElement("textarea");
          textArea.value = text;
          textArea.style.position = "fixed";
          textArea.style.left = "-9999px";
          document.body.appendChild(textArea);
          textArea.focus();
          textArea.select();
          document.execCommand("copy");
          document.body.removeChild(textArea);
        }

        button.textContent = "Copied!";
        setTimeout(() => {
          button.textContent = original;
        }, 1800);
      } catch (err) {
        button.textContent = "Copy failed";
        setTimeout(() => {
          button.textContent = original;
        }, 1800);
      }
    }
  </script>
</body>
</html>