File size: 35,548 Bytes
d00556c
193ee10
d00556c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e9628be
d00556c
 
e9628be
d00556c
 
e9628be
d00556c
 
 
 
 
 
 
 
 
 
 
 
e9628be
d00556c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
785f414
 
d00556c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
785f414
 
d00556c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8d50cf2
d00556c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e9628be
 
 
 
 
 
 
 
 
 
d00556c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
193ee10
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
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
<!DOCTYPE html>
<html>
<head>
  <meta charset="utf-8">
  <meta name="description"
        content="Deformable Neural Radiance Fields creates free-viewpoint portraits (nerfies) from casually captured videos.">
  <meta name="keywords" content="Nerfies, D-NeRF, NeRF">
  <meta name="viewport" content="width=device-width, initial-scale=1">
  <title>CarBoN: Calibrated Best-of-N Sampling Improves Test-time Reasoning</title>

  <!-- Global site tag (gtag.js) - Google Analytics -->
  <script async src="https://www.googletagmanager.com/gtag/js?id=G-PYVRSFMDRL"></script>
  <script>
    window.dataLayer = window.dataLayer || [];

    function gtag() {
      dataLayer.push(arguments);
    }

    gtag('js', new Date());

    gtag('config', 'G-PYVRSFMDRL');
  </script>

  <link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro"
        rel="stylesheet">

  <link rel="stylesheet" href="./static/css/bulma.min.css">
  <link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
  <link rel="stylesheet" href="./static/css/bulma-slider.min.css">
  <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">
  <link rel="stylesheet" href="./static/css/index.css">
  <link rel="icon" href="./static/images/favicon.svg">

  <script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
  <script defer src="./static/js/fontawesome.all.min.js"></script>
  <script src="./static/js/bulma-carousel.min.js"></script>
  <script src="./static/js/bulma-slider.min.js"></script>
  <script src="./static/js/index.js"></script>
  
  <!-- MathJax for mathematical notation -->
  <script src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>
  <script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
</head>
<body>

<section class="hero">
  <div class="hero-body">
    <div class="container is-max-desktop">
      <div class="columns is-centered">
        <div class="column has-text-centered">
          <h1 class="title is-1 publication-title">CarBoN: Calibrated Best-of-N Sampling<br>Improves Test-time Reasoning</h1>
          <div class="is-size-5 publication-authors">
            <span class="author-block">
              <a href="https://sites.google.com/view/yungchentang" target="_blank" rel="noopener noreferrer">Yung-Chen Tang</a><sup>1,2</sup>,
            </span>
            <span class="author-block">
              <a href="https://sites.google.com/site/pinyuchenpage/home" target="_blank" rel="noopener noreferrer">Pin-Yu Chen</a><sup>3</sup>,
            </span>
            <span class="author-block">
              <a href="https://people.epfl.ch/andrea.cavallaro?lang=en" target="_blank" rel="noopener noreferrer">Andrea Cavallaro</a><sup>1,2</sup>
            </span>
          </div>
          <div class="is-size-5 publication-authors">
            <span class="author-block"><sup>1</sup>EPFL,</span>
            <span class="author-block"><sup>2</sup>Idiap Research Institute,</span>
            <span class="author-block"><sup>3</sup>IBM Research</span>
          </div>

          <div class="column has-text-centered">
            <div class="publication-links">
              <!-- arXiv Link. -->
              <span class="link-block">
                <a href="https://arxiv.org/abs/2510.15674" target="_blank" rel="noopener noreferrer"
                   class="external-link button is-normal is-rounded is-dark">
                  <span class="icon">
                      <i class="ai ai-arxiv"></i>
                  </span>
                  <span>arXiv</span>
                </a>
              </span>
              <!-- Code Link. -->
              <span class="link-block">
                <a href="#"
                   class="external-link button is-normal is-rounded is-dark">
                  <span class="icon">
                      <i class="fab fa-github"></i>
                  </span>
                  <span>Code</span>
                  </a>
              </span>
            </div>

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


<section class="section">
  <div class="container is-max-desktop">
    <!-- Abstract. -->
    <div class="columns is-centered has-text-centered">
      <div class="column is-four-fifths">
        <h2 class="title is-3">Abstract</h2>
        <div class="content has-text-justified">
          <p>
            Allocating more computation during inference time (test-time scaling) improves language model performance, especially for reasoning tasks.
            However, popular methods like Best-of-N sampling often show diminishing returns as N increases.
            To address this inefficiency, we introduce a general <strong>test-time calibration framework</strong> that adaptively modifies the model toward high-reward reasoning paths, with theoretical guarantees of improving the lower bound of expected reward under finite sampling, all without large language model (LLM) retraining.
            Within this framework, we propose <strong>CarBoN</strong> (Calibrated Best-of-N), a two-phase method that first explores the solution space and then learns a calibration of the logits via an input-specific temperature T and additive shift vector δ, guiding generation toward more reliable reasoning.
          </p>
          <p>
            Experiments on MATH-500 and AIME-2024 show that CarBoN improves efficiency, with up to 4× fewer rollouts to reach the same accuracy, while often achieving higher accuracy under fixed budgets.
            We also analyze the complementary roles of T and δ in balancing output diversity and correctness, and demonstrate that the framework also generalizes to step-level sampling strategies such as beam search.
          </p>
        </div>
      </div>
    </div>
    <!--/ Abstract. -->

  </div>
</section>


<section class="section">
  <div class="container is-max-desktop">

    <!-- Motivation. -->
    <div class="columns is-centered">
      <div class="column is-full-width">
        <h2 class="title is-3">Why calibration for test-time scaling (TTS)? A motivating example of reward-based binary search.</h2>

        <div class="content has-text-justified">
          <p>
            Let the task be finding a target in [0,10<sup>4</sup>]. 
            Calibration means that before each search step the model can query <i>n</i> candidate points for reward, where <i>n</i> denotes the number of reward queries per step.
            The reward is defined as the inverse distance to the target plus noise.
            The baseline binary search, corresponding to naive TTS (<i>n</i>=0), requires 13.3 steps on average. Increasing <i>n</i> significantly accelerates convergence: for example, with <i>n</i>=16 the search depth is reduced by up to 74% (see Figure 1, left).
            Figure 1 (right) shows an example run where calibration quickly converges to the target, while vanilla binary search continues oscillating.
            This example highlights that reward feedback for calibration reshapes the sampling distribution and motivates its use for TTS.
          </p>
        </div>

        <!-- Figure 1 -->
        <div class="content has-text-centered">
          <img src="./static/images/figure_1.png" alt="Reward-guided calibration accelerates binary search" style="width: 100%; height: auto; margin: 20px 0; display: block;">
          <p class="has-text-justified" style="margin: 0;">
            <strong>Figure 1. Reward-guided calibration accelerates binary search.</strong>
            Left: Increasing per-step noisy reward (inverse-distance signal + noise) lowers average search steps versus vanilla.
            Right: Example showing reward guidance converges early; vanilla keeps oscillating.
          </p>
        </div>
        <!--/ Figure 1 -->

      </div>
    </div>
    <!--/ Motivation. -->


  </div>
</section>


<section class="section">
  <div class="container is-max-desktop">

    <!-- Framework Overview. -->
    <div class="columns is-centered">
      <div class="column is-full-width">
        <h2 class="title is-3">Overview of Test-time Calibration Framework</h2>

        <div class="content has-text-justified">
          <p>
            Building on this principle, our framework reuses sampled completions that are normally discarded in parallel sampling methods to extract reward signals and perform calibration.
            Within this framework, we introduce <strong>CarBoN</strong> (Calibrated Best-of-N). Without modifying the original LLM, our framework allocates part of the budget to exploration and calibration, then focuses the remaining budget on high-scoring regions using logit calibration.
            Reusing high-scoring answer selected by reward model enhances answer quality and query efficiency, under the same inference budget.
          </p>
        </div>

        <!-- Figure 2 -->
        <div class="content has-text-centered">
          <img src="./static/images/figure_2.png" alt="Test-time calibration framework and MATH-500 results" style="width: 100%; height: auto; margin: 20px 0; display: block;">
          <p class="has-text-justified" style="margin: 0;">
            <strong>Figure 2. (a) Test-time calibration framework.</strong> With a rollout budget N = N<sub>1</sub> + N<sub>2</sub>, the model first explores by generating and scoring N<sub>1</sub> candidate responses. The model then learns calibration parameters (δ, T) from high-scoring responses, using them to adjust the logits for the remaining N<sub>2</sub> generations. The final answer is selected from all N candidates.
            <strong>(b) MATH-500 Results.</strong> CarBoN improves weighted Best-of-N accuracy across four models. For all models, calibrated accuracy at N=64 (orange dash line) matches or exceeds uncalibrated accuracy at N=256, corresponding to up to a 4× reduction in rollout budgets. Notably, with Qwen2.5-Math-1.5B-Instruct at N=64, CarBoN surpasses GPT-4o (red dashed line), while uncalibrated Best-of-N with N=256 does not.
          </p>
        </div>
        <!--/ Figure 2 -->

      </div>
    </div>
    <!--/ Framework Overview. -->


  </div>
</section>


<section class="section">
  <div class="container is-max-desktop">

    <!-- Theoretical Guarantees. -->
    <div class="columns is-centered">
      <div class="column is-full-width">
        <h2 class="title is-3">Theoretical Analysis</h2>

        <div class="content has-text-justified">
          <p>
            We provide theoretical foundations for our calibration framework through three formal results. 
            <strong>Lemma 1</strong> establishes the existence of joint calibration parameters \((\delta^*, T^*)\) that provably improve the model's predictive distribution when the base model is not perfectly calibrated. 
            <strong>Theorem 1</strong> proves that applying such calibration strictly increases the expected reward under finite Best-of-\(N\) sampling, by showing that calibration achieves first-order stochastic dominance—making high-reward outputs more likely. 
            <strong>Corollary 1</strong> demonstrates that discarding exploration samples and using only calibrated samples is strictly suboptimal. 
            Together, these results rigorously justify why test-time calibration works and why both phases are essential.
          </p>
        </div>

        <!-- Lemma 1 -->
        <details style="margin-bottom: 1rem; border: 1px solid #dbdbdb; border-radius: 6px; padding: 1rem;">
          <summary style="cursor: pointer; font-weight: bold; font-size: 1.1rem; color: #3273dc;">
            Lemma 1: Existence of an Improving Joint Solution \((\delta, T)\)
          </summary>
          <div style="margin-top: 1rem; padding-left: 1rem; border-left: 3px solid #3273dc;">
            <p>
              Let the joint loss function be \(\mathcal{L}(\delta, T) = \mathbb{E}_{y \sim \mathcal{D}_\text{calib}(x)} \left[ -\log p_\theta(y \mid x; \delta, T) \right]\).
              Let \(\bar{p}_{\theta}\) be the model's average predictive distribution and \(\bar{p}_{\text{target}}\) be the empirical average one-hot distribution, both averaged over all generation steps in the calibration set \(\mathcal{D}_\text{calib}(x)\).
              Suppose the base model is not perfectly calibrated in the sense that at least one of the following conditions holds: (1) \(\bar{p}_{\theta} \neq \bar{p}_{\text{target}}\), or (2) the average logit of ground-truth tokens does not equal the average expected logit.
            </p>
            <p>
              Then there exists a joint solution \((\delta, T) \in \mathbb{R}^D \times (0, \infty)\), where \((\delta, T) \neq (\mathbf{0}, 1)\), such that the loss is strictly reduced: 
              \[\mathcal{L}(\delta, T) < \mathcal{L}(\mathbf{0}, 1)\]
            </p>
          </div>
        </details>

        <!-- Theorem 1 -->
        <details style="margin-bottom: 1rem; border: 1px solid #dbdbdb; border-radius: 6px; padding: 1rem;">
          <summary style="cursor: pointer; font-weight: bold; font-size: 1.1rem; color: #3273dc;">
            Theorem 1: Joint Calibration \((\delta, T)\) Improves Expected Reward from Best-of-\(N\) Sampling
          </summary>
          <div style="margin-top: 1rem; padding-left: 1rem; border-left: 3px solid #3273dc;">
            <p>
              Let \(p_{\theta}(y \mid x; \delta, T)\) be the model's probability distribution over outputs \(y \in \mathcal{Y}\), parameterized by a calibration vector \(\delta\) and a temperature \(T\). Let the base model be configured with parameters \((\mathbf{0}, T_{base})\) for some \(T_{base} > 0\).
              Let \(R(x,y)\) be a reward function, and assume there exists a unique output \(y^* \in \mathcal{Y}\) with a strictly maximum reward, i.e., \(r^* = R(x,y^*) > \max_{y \neq y^*} R(x,y) = r_{\text{other\_max}}\).
            </p>
            <p>
              We consider cases where joint calibration with parameters \((\delta^*, T^*)\) improves upon the base model by increasing the probability of the unique optimal output:
              \[p_{\theta}(y^* \mid x; \delta^*, T^*) > p_{\theta}(y^* \mid x; \mathbf{0}, T_{base})\]
            </p>
            <p>
              Then, for any \(n \geq 1\) within the remaining inference budget after calibration, the lower bound on the expected best-of-\(N\) reward under the jointly calibrated model is strictly greater than that of the base model.
              Specifically, let \(R_{LB}(p) = r^* - (1-p)^n (r^* - r_{\text{other\_max}})\) be a valid lower bound for the expected best-of-\(N\) reward, where \(p\) is the probability of sampling \(y^*\). The improvement in this lower bound is strictly positive:
              \[\Delta_{R_{LB}}(x,n) = R_{LB}(p_{\theta}(y^* \mid x; \delta^*, T^*)) - R_{LB}(p_{\theta}(y^* \mid x; \mathbf{0}, T_{base})) > 0\]
            </p>
          </div>
        </details>

        <!-- Corollary 1 -->
        <details style="margin-bottom: 1rem; border: 1px solid #dbdbdb; border-radius: 6px; padding: 1rem;">
          <summary style="cursor: pointer; font-weight: bold; font-size: 1.1rem; color: #3273dc;">
            Corollary 1: Sub-optimality of Exploitation Alone
          </summary>
          <div style="margin-top: 1rem; padding-left: 1rem; border-left: 3px solid #3273dc;">
            <p>
              The final candidate is selected by maximizing \(R(x, y)\) over a set of candidates \(\mathcal{Y}\). Since \(\mathcal{Y}_{\text{exploit}}\) is a subset of the union \(\mathcal{Y} = \mathcal{Y}_{\text{explore}} \cup \mathcal{Y}_{\text{exploit}}\), the strategy of only selecting from \(\mathcal{Y}_{\text{exploit}}\) is sub-optimal compared to selecting from the union.
            </p>
            <p>
              This is because the maximum reward achievable from the union is greater than or equal to the maximum reward achievable from the exploitation set alone:
              \[\max_{y \in \mathcal{Y}_{\text{explore}} \cup \mathcal{Y}_{\text{exploit}}} R(x, y) \geq \max_{y \in \mathcal{Y}_{\text{exploit}}} R(x, y)\]
            </p>
            <p>
              Therefore, including exploration samples in the final candidate pool is essential for maximizing expected reward.
            </p>
          </div>
        </details>

      </div>
    </div>
    <!--/ Theoretical Guarantees. -->


  </div>
</section>


<section class="section">
  <div class="container is-max-desktop">

    <!-- Empirical Results. -->
    <div class="columns is-centered">
      <div class="column is-full-width">
        <h2 class="title is-3">Empirical Results</h2>

        <div class="content has-text-justified">
          <p>
            We evaluate CarBoN on MATH-500 and AIME-2024 benchmarks across multiple language models. The results demonstrate that CarBoN consistently improves upon standard Best-of-\(N\) sampling, with calibrated accuracy at \(N=64\) matching or exceeding uncalibrated results at \(N=256\) (\(4\times\) reduction in rollout budget).
          </p>
        </div>

        <!-- Table 1: MATH-500 Results -->
        <h3 class="title is-4" style="margin-top: 2rem;">MATH-500 Results</h3>
        <div style="max-width: 800px; margin: 0 auto 2rem auto;">
          <p style="font-size: 0.95rem; margin-bottom: 0.5rem;">
            <strong>Table 1.</strong> Accuracy (%) of four models on MATH-500, comparing Weighted Best-of-<span style='font-style:italic;'>N</span> methods before and after calibration. CarBoN enables further improvements beyond the plateau of standard Best-of-<span style='font-style:italic;'>N</span>, with calibrated accuracy at <span style='font-style:italic;'>N</span>=64 exceeding the uncalibrated results at <span style='font-style:italic;'>N</span>=256. Bold indicates better accuracy for each <span style='font-style:italic;'>N</span>.
          </p>
          <div style="overflow-x: auto;">
            <table class="table is-bordered is-striped is-narrow is-hoverable" style="margin: 0 auto; width: 100%; min-width: 600px;">
              <thead>
                <tr>
                  <th style="text-align: center; width: 180px;">Model</th>
                  <th style="text-align: center; width: 110px;">Method</th>
                  <th style="text-align: center; width: 70px;">N=8</th>
                  <th style="text-align: center; width: 70px;">N=16</th>
                  <th style="text-align: center; width: 70px;">N=32</th>
                  <th style="text-align: center; width: 70px;">N=64</th>
                  <th style="text-align: center; width: 70px;">N=128</th>
                  <th style="text-align: center; width: 70px;">N=256</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: white;">Llama-3.2-1B-Instruct</td>
                  <td style="text-align: left; background-color: white;">Best-of-<span style='font-style:italic;'>N</span></td>
                  <td style="text-align: center; background-color: white;">42.0</td>
                  <td style="text-align: center; background-color: white;">44.6</td>
                  <td style="text-align: center; background-color: white;">47.8</td>
                  <td style="text-align: center; background-color: white;">48.6</td>
                  <td style="text-align: center; background-color: white;">50.6</td>
                  <td style="text-align: center; background-color: white;">50.8</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: white;">CarBoN</td>
                  <td style="text-align: center; background-color: white;"><strong>43.0</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>45.6</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>48.4</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>51.0</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>51.8</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>51.8</strong></td>
                </tr>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: #f9f9f9;">Llama-3.2-3B-Instruct</td>
                  <td style="text-align: left; background-color: #f9f9f9;">Best-of-<span style='font-style:italic;'>N</span></td>
                  <td style="text-align: center; background-color: #f9f9f9;">56.8</td>
                  <td style="text-align: center; background-color: #f9f9f9;">58.2</td>
                  <td style="text-align: center; background-color: #f9f9f9;">59.6</td>
                  <td style="text-align: center; background-color: #f9f9f9;">61.6</td>
                  <td style="text-align: center; background-color: #f9f9f9;">61.8</td>
                  <td style="text-align: center; background-color: #f9f9f9;">62.2</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: #f9f9f9;">CarBoN</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>57.6</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>59.0</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>60.8</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>62.2</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>63.2</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>63.4</strong></td>
                </tr>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: white;">Qwen2.5-1.5B-Instruct</td>
                  <td style="text-align: left; background-color: white;">Best-of-<span style='font-style:italic;'>N</span></td>
                  <td style="text-align: center; background-color: white;"><strong>56.4</strong></td>
                  <td style="text-align: center; background-color: white;">57.6</td>
                  <td style="text-align: center; background-color: white;">61.4</td>
                  <td style="text-align: center; background-color: white;">62.0</td>
                  <td style="text-align: center; background-color: white;">62.6</td>
                  <td style="text-align: center; background-color: white;">62.2</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: white;">CarBoN</td>
                  <td style="text-align: center; background-color: white;">55.0</td>
                  <td style="text-align: center; background-color: white;"><strong>60.0</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>61.8</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>62.4</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>64.0</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>64.4</strong></td>
                </tr>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: #f9f9f9;">Qwen2.5-Math-1.5B-Instruct</td>
                  <td style="text-align: left; background-color: #f9f9f9;">Best-of-<span style='font-style:italic;'>N</span></td>
                  <td style="text-align: center; background-color: #f9f9f9;">73.6</td>
                  <td style="text-align: center; background-color: #f9f9f9;">75.4</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>76.4</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;">75.6</td>
                  <td style="text-align: center; background-color: #f9f9f9;">76.4</td>
                  <td style="text-align: center; background-color: #f9f9f9;">76.8</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: #f9f9f9;">CarBoN</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>74.2</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>76.0</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>76.4</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>77.2</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>77.2</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>77.8</strong></td>
                </tr>
              </tbody>
            </table>
          </div>
        </div>

        <!-- Table 2: AIME-2024 Results -->
        <h3 class="title is-4" style="margin-top: 3rem;">AIME-2024 Results</h3>
        <div style="max-width: 800px; margin: 0 auto 2rem auto;">
          <p style="font-size: 0.95rem; margin-bottom: 0.5rem;">
            <strong>Table 2.</strong> Correct answers (out of 30) on the AIME-2024 benchmark for two math-specialized models, comparing Best-of-<span style='font-style:italic;'>N</span> and CarBoN across different rollout budgets. CarBoN enables further improvements beyond the plateau of standard Best-of-<span style='font-style:italic;'>N</span>. Bold numbers indicate the higher number of correct answers for each <span style='font-style:italic;'>N</span>.
          </p>
          <div style="overflow-x: auto;">
            <table class="table is-bordered is-striped is-narrow is-hoverable" style="margin: 0 auto; width: 100%; min-width: 600px;">
              <thead>
                <tr>
                  <th style="text-align: center; width: 180px;">Model</th>
                  <th style="text-align: center; width: 110px;">Method</th>
                  <th style="text-align: center; width: 70px;">N=16</th>
                  <th style="text-align: center; width: 70px;">N=32</th>
                  <th style="text-align: center; width: 70px;">N=64</th>
                  <th style="text-align: center; width: 70px;">N=128</th>
                  <th style="text-align: center; width: 70px;">N=256</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: white;">Qwen2.5-Math-1.5B-Instruct</td>
                  <td style="text-align: left; background-color: white;">Best-of-<span style='font-style:italic;'>N</span></td>
                  <td style="text-align: center; background-color: white;">4/30</td>
                  <td style="text-align: center; background-color: white;">5/30</td>
                  <td style="text-align: center; background-color: white;">6/30</td>
                  <td style="text-align: center; background-color: white;">6/30</td>
                  <td style="text-align: center; background-color: white;">6/30</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: white;">CarBoN</td>
                  <td style="text-align: center; background-color: white;">4/30</td>
                  <td style="text-align: center; background-color: white;">5/30</td>
                  <td style="text-align: center; background-color: white;">6/30</td>
                  <td style="text-align: center; background-color: white;"><strong>7/30</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>7/30</strong></td>
                </tr>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: #f9f9f9;">Qwen2.5-Math-7B-Instruct</td>
                  <td style="text-align: left; background-color: #f9f9f9;">Best-of-<span style='font-style:italic;'>N</span></td>
                  <td style="text-align: center; background-color: #f9f9f9;">5/30</td>
                  <td style="text-align: center; background-color: #f9f9f9;">5/30</td>
                  <td style="text-align: center; background-color: #f9f9f9;">6/30</td>
                  <td style="text-align: center; background-color: #f9f9f9;">6/30</td>
                  <td style="text-align: center; background-color: #f9f9f9;">6/30</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: #f9f9f9;">CarBoN</td>
                  <td style="text-align: center; background-color: #f9f9f9;">5/30</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>6/30</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9">6/30</td>
                  <td style="text-align: center; background-color: #f9f9f9">6/30</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>7/30</strong></td>
                </tr>
              </tbody>
            </table>
          </div>
        </div>

        <!-- Generalizing Test-time Calibration Beyond Best-of-N -->
        <h3 class="title is-4" style="margin-top: 3rem;">Generalizing Test-time Calibration Beyond Best-of-<span style='font-style:italic;'>N</span></h3>
        <div class="content has-text-justified" style="margin-bottom: 1rem;">
          <p>
            Test-time calibration is not limited to Best-of-<span style='font-style:italic;'>N</span> sampling. We show that calibrating step-level strategies such as beam search can further improve sample efficiency and accuracy. As shown below, calibrated beam search achieves similar or better performance with fewer candidates, demonstrating the generality of our approach.
          </p>
        </div>
        <div style="max-width: 800px; margin: 0 auto 2rem auto;">
          <p style="font-size: 0.95rem; margin-bottom: 0.5rem;">
            <strong>Table 3.</strong> Accuracy (%) of standard and calibrated beam search on the MATH-500 benchmark. Calibrated beam search generally improves test-time reasoning performance, especially for larger <span style='font-style:italic;'>N</span>.
          </p>
          <div style="overflow-x: auto;">
            <table class="table is-bordered is-striped is-narrow is-hoverable" style="margin: 0 auto; width: 100%; min-width: 600px;">
              <thead>
                <tr>
                  <th style="text-align: center; width: 180px;">Model</th>
                  <th style="text-align: center; width: 170px;">Method</th>
                  <th style="text-align: center; width: 70px;">N=8</th>
                  <th style="text-align: center; width: 70px;">N=16</th>
                  <th style="text-align: center; width: 70px;">N=32</th>
                  <th style="text-align: center; width: 70px;">N=64</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: white;">Llama-3.2-1B-Instruct</td>
                  <td style="text-align: left; background-color: white;">Beam Search</td>
                  <td style="text-align: center; background-color: white;">56.0</td>
                  <td style="text-align: center; background-color: white;">58.4</td>
                  <td style="text-align: center; background-color: white;">60.4</td>
                  <td style="text-align: center; background-color: white;">62.2</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: white;">Calibrated Beam Search</td>
                  <td style="text-align: center; background-color: white;"><strong>57.2</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>60.0</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>62.2</strong></td>
                  <td style="text-align: center; background-color: white;"><strong>64.2</strong></td>
                </tr>
                <tr>
                  <td rowspan="2" style="vertical-align: middle; text-align: left; background-color: #f9f9f9;">Qwen2.5-Math-1.5B-Instruct</td>
                  <td style="text-align: left; background-color: #f9f9f9;">Beam Search</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>79.0</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;">79.2</td>
                  <td style="text-align: center; background-color: #f9f9f9;">80.2</td>
                  <td style="text-align: center; background-color: #f9f9f9;">81.4</td>
                </tr>
                <tr>
                  <td style="text-align: left; background-color: #f9f9f9;">Calibrated Beam Search</td>
                  <td style="text-align: center; background-color: #f9f9f9;">78.6</td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>79.6</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>81.2</strong></td>
                  <td style="text-align: center; background-color: #f9f9f9;"><strong>82.8</strong></td>
                </tr>
              </tbody>
            </table>
          </div>
        </div>
        <div class="content has-text-justified" style="margin-top: 2rem;">
          <p>
            For more detailed ablation studies and analysis, please refer to our paper.
          </p>
        </div>

      </div>
    </div>
    <!--/ Empirical Results. -->

  </div>
</section>


<section class="section" id="BibTeX">
  <div class="container is-max-desktop content">
    <h2 class="title">BibTeX</h2>
    <pre><code>@misc{tang2025carboncalibratedbestofnsampling,
  title={CarBoN: Calibrated Best-of-N Sampling Improves Test-time Reasoning}, 
  author={Yung-Chen Tang and Pin-Yu Chen and Andrea Cavallaro},
  year={2025},
  eprint={2510.15674},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2510.15674}, 
}
</code></pre>
  </div>
</section>


<footer class="footer">
  <div class="container">
    <div class="columns is-centered">
      <div class="column is-8">
        <div class="content">
          <p>
            This website is licensed under a <a rel="license"
                                                href="http://creativecommons.org/licenses/by-sa/4.0/">Creative
            Commons Attribution-ShareAlike 4.0 International License</a>.
          </p>
          <p>
            This means you are free to borrow the <a
              href="https://github.com/nerfies/nerfies.github.io">source code</a> of this website,
            we just ask that you link back to this page in the footer.
            Please remember to remove the analytics code included in the header of the website which
            you do not want on your website.
          </p>
        </div>
      </div>
    </div>
  </div>
</footer>

</body>
</html>