terms-bench.github.io/submit.html at main · Terms-bench/terms-bench.github.io · 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
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
  <title>TERMS-Bench: Submit your model</title>
  <script>
    // Resolve the theme before first paint so we never flash the wrong palette.
    (function () {
      try {
        var saved = localStorage.getItem('terms-theme');
        var theme = saved || (
          window.matchMedia &&
          window.matchMedia('(prefers-color-scheme: dark)').matches
            ? 'dark' : 'light'
        );
        document.documentElement.setAttribute('data-theme', theme);
      } catch (_) {
        document.documentElement.setAttribute('data-theme', 'light');
      }
    })();
  </script>
  <link rel="stylesheet" href="style.css" />
</head>
<body>
  <button class="theme-toggle" id="theme-toggle" type="button"
          aria-label="Toggle color theme" title="Toggle light / dark">
    <svg class="theme-icon theme-sun" viewBox="0 0 24 24" aria-hidden="true">
      <circle cx="12" cy="12" r="4" fill="currentColor"></circle>
      <g stroke="currentColor" stroke-width="1.6" stroke-linecap="round" fill="none">
        <line x1="12" y1="2.5" x2="12" y2="5"></line>
        <line x1="12" y1="19" x2="12" y2="21.5"></line>
        <line x1="2.5" y1="12" x2="5" y2="12"></line>
        <line x1="19" y1="12" x2="21.5" y2="12"></line>
        <line x1="4.8" y1="4.8" x2="6.6" y2="6.6"></line>
        <line x1="17.4" y1="17.4" x2="19.2" y2="19.2"></line>
        <line x1="4.8" y1="19.2" x2="6.6" y2="17.4"></line>
        <line x1="17.4" y1="6.6" x2="19.2" y2="4.8"></line>
      </g>
    </svg>
    <svg class="theme-icon theme-moon" viewBox="0 0 24 24" aria-hidden="true">
      <path d="M20 14.5A8.5 8.5 0 0 1 9.5 4 8.5 8.5 0 1 0 20 14.5z"
            fill="currentColor"></path>
    </svg>
  </button>

  <main class="page page-submit">

    <header class="masthead masthead-fingerprints">
      <div class="masthead-row">
        <h1 class="wordmark"><a class="wordmark-link" href="index.html">TERMS-Bench</a></h1>
        <p class="page-eyebrow">Submit your model</p>
      </div>
      <p class="page-lede">
        Want your model on the TERMS-Bench leaderboard? You provide the
        API access; we run the full diagnostic suite on our backend and
        publish the results. No local infra, no eval harness setup,
        no GPU. Submissions typically run within 3-5 business days.
      </p>
      <nav class="fp-back-nav" aria-label="Back to leaderboard">
        <a href="index.html" class="fp-back-link">&larr; Back to leaderboard</a>
      </nav>
    </header>

    <!-- =========================================================== CTA -->
    <section class="submit-cta">
      <div class="submit-cta-card">
        <h2 class="submit-cta-title">One submission, full diagnostic report</h2>
        <p class="submit-cta-body">
          Each accepted submission is evaluated on the full paper suite:
          the <code>3 &times; 6</code> regime &times; family design matrix
          across the bilateral price negotiation protocol. You receive
          all four diagnostic axes (<code>SE&#8314;</code>,
          <code>AGR&#8314;</code>, <code>FAGR&#8315;</code>,
          <code>BE_type</code>, <code>CritViol%</code>), per-family
          stratification, per-regime breakdowns, and a bargaining
          fingerprint card. Cost is covered by your API key; we publish
          aggregate spend so you know what to expect.
        </p>
        <p class="submit-cta-note">
          Submissions are emailed to
          <a href="mailto:submissions@terms-bench.org">submissions@terms-bench.org</a>.
          For the API key we recommend a
          <a href="https://onetimesecret.com" target="_blank" rel="noopener">
          one-time secret link</a> (paste it in the notes field) so the
          key never sits in your sent-mail archive. We use the key only
          for the eval run, never log it, and recommend you create a
          dedicated, scope-limited key with a hard spend cap.
        </p>
      </div>
    </section>

    <!-- ===================================================== How it works -->
    <h2 class="submit-section-h">How submission works</h2>
    <ol class="submit-steps">
      <li>
        <h3>1. Pick a provider</h3>
        <p>
          Any OpenAI-compatible chat-completions endpoint works:
          OpenAI, Anthropic (via the dedicated route), OpenRouter,
          Together, Fireworks, Groq, or your own self-hosted vLLM /
          SGLang server with the OpenAI-compatible adapter.
        </p>
      </li>
      <li>
        <h3>2. Mint a dedicated API key</h3>
        <p>
          We recommend creating a fresh key with a hard spend cap
          (typically $50-$200 covers the full paper suite, depending on
          model). The key is transported over TLS, used by our eval
          runner only, and never persisted to disk on our side.
        </p>
      </li>
      <li>
        <h3>3. Fill in the form and send the email</h3>
        <p>
          Fill in model identifier (e.g. <code>openai/gpt-5</code>,
          <code>anthropic/claude-opus-4-7</code>,
          <code>meta-llama/llama-4-70b-instruct</code>), provider, your
          contact info, and any sampling notes. Hitting submit composes
          an email to
          <a href="mailto:submissions@terms-bench.org">submissions@terms-bench.org</a>
          with everything pre-filled - review and send from your mail
          client. For the API key, paste a one-time secret link in the
          notes field (recommended) or include it directly if you accept
          the risk.
        </p>
      </li>
      <li>
        <h3>4. We run the eval, you get the report</h3>
        <p>
          We run the full paper suite (single replicate, paper-aligned
          cell sizes). Within 3-5 business days, you receive a private
          report with all metrics, traces of representative episodes,
          and the leaderboard PR for your sign-off before publishing.
        </p>
      </li>
    </ol>

    <!-- ===================================================== The form -->
    <h2 class="submit-section-h">Submission</h2>
    <form id="submit-form" class="submit-form" autocomplete="off" novalidate>
      <div class="submit-form-row">
        <div class="submit-field">
          <label for="f-name" class="submit-field-label">
            Your name<span class="submit-required">*</span>
          </label>
          <input id="f-name" name="submitter_name" class="submit-input"
                 type="text" required maxlength="120" />
        </div>
        <div class="submit-field">
          <label for="f-email" class="submit-field-label">
            Contact email<span class="submit-required">*</span>
          </label>
          <input id="f-email" name="submitter_email" class="submit-input"
                 type="email" required maxlength="200" />
        </div>
      </div>

      <div class="submit-field">
        <label for="f-affiliation" class="submit-field-label">
          Affiliation
        </label>
        <input id="f-affiliation" name="affiliation" class="submit-input"
               type="text" maxlength="200" placeholder="Optional, shown on leaderboard" />
      </div>

      <div class="submit-form-row">
        <div class="submit-field">
          <label for="f-display" class="submit-field-label">
            Model display name<span class="submit-required">*</span>
          </label>
          <input id="f-display" name="display_name" class="submit-input"
                 type="text" required maxlength="80"
                 placeholder="e.g. Claude Opus 4.8" />
          <span class="submit-field-hint">
            Shown on the leaderboard exactly as written.
          </span>
        </div>
        <div class="submit-field">
          <label for="f-model" class="submit-field-label">
            Model identifier<span class="submit-required">*</span>
          </label>
          <input id="f-model" name="model_id" class="submit-input"
                 type="text" required maxlength="200"
                 placeholder="e.g. anthropic/claude-opus-4-8" />
          <span class="submit-field-hint">
            The exact string passed to the provider's API.
          </span>
        </div>
      </div>

      <div class="submit-form-row">
        <div class="submit-field">
          <label for="f-provider" class="submit-field-label">
            Provider<span class="submit-required">*</span>
          </label>
          <select id="f-provider" name="provider" class="submit-select" required>
            <option value="">Select a provider</option>
            <option value="openai">OpenAI</option>
            <option value="anthropic">Anthropic</option>
            <option value="openrouter">OpenRouter</option>
            <option value="together">Together AI</option>
            <option value="fireworks">Fireworks AI</option>
            <option value="groq">Groq</option>
            <option value="custom">Custom (OpenAI-compatible)</option>
          </select>
        </div>
        <div class="submit-field">
          <label for="f-base-url" class="submit-field-label">
            API base URL
          </label>
          <input id="f-base-url" name="base_url" class="submit-input"
                 type="url" maxlength="300"
                 placeholder="Auto-detected from provider; override only if needed" />
          <span class="submit-field-hint">
            Required only for <code>custom</code> / self-hosted endpoints.
          </span>
        </div>
      </div>

      <div class="submit-field">
        <label for="f-notes" class="submit-field-label">
          Notes &mdash; API key delivery, sampling, reasoning
          <span class="submit-required">*</span>
        </label>
        <textarea id="f-notes" name="notes" class="submit-textarea"
                  required maxlength="2000" rows="5"
                  placeholder='Recommended for the API key: paste a one-time secret link from https://onetimesecret.com or 1Password / Bitwarden Send.

Optional sampling overrides:
  {"temperature": 0.0, "reasoning_effort": "medium"}'></textarea>
        <span class="submit-field-hint">
          We default to <code>temperature=0.0</code> and the model's
          standard reasoning mode unless you say otherwise. Recommended
          API-key delivery: paste a one-time secret URL here (e.g. from
          <a href="https://onetimesecret.com" target="_blank" rel="noopener">onetimesecret.com</a>)
          so the raw key never lives in your sent-mail archive.
        </span>
      </div>

      <label class="submit-consent">
        <input id="f-consent" name="consent" type="checkbox" required />
        <span>
          I authorize TERMS-Bench to run the full paper suite using my
          API access (typical cost <code>$50-$200</code>, charged to my
          account), and to publish aggregate results on the public
          leaderboard. I will receive a private report and the
          leaderboard PR for sign-off before publication.
        </span>
      </label>

      <div class="submit-actions">
        <button type="submit" id="submit-btn" class="submit-submit-button">
          Compose submission email
        </button>
        <span id="submit-status" class="submit-status" role="status" aria-live="polite"></span>
      </div>

      <p class="submit-field-hint" style="margin-top:8px;">
        Prefer your own email client? Just email
        <a href="mailto:submissions@terms-bench.org">submissions@terms-bench.org</a>
        with the model identifier, provider, your contact, and a
        one-time secret link for the API key.
      </p>
    </form>

    <!-- ===================================================== Providers -->
    <h2 class="submit-section-h">Supported providers</h2>
    <section class="submit-providers">
      <p class="submit-providers-lede">
        Any of these work out of the box. <code>custom</code> covers
        anything OpenAI-compatible (vLLM, SGLang, Ollama with the OpenAI
        adapter, self-hosted endpoints).
      </p>
      <div class="provider-grid">
        <div class="provider-card">
          <h3 class="provider-name">OpenAI</h3>
          <span class="provider-url">https://api.openai.com/v1</span>
          <span class="provider-key">Header: <code>Authorization: Bearer sk-...</code></span>
          <p class="provider-note">
            Reasoning models (<code>o1</code>, <code>o3</code>,
            <code>gpt-5</code> family) handled via the reasoning-effort
            param; pass <code>reasoning_effort</code> in notes.
          </p>
        </div>
        <div class="provider-card">
          <h3 class="provider-name">Anthropic</h3>
          <span class="provider-url">https://api.anthropic.com/v1</span>
          <span class="provider-key">Header: <code>x-api-key: sk-ant-...</code></span>
          <p class="provider-note">
            Native client; thinking models (<code>claude-opus-4.x</code>)
            run with extended thinking by default.
          </p>
        </div>
        <div class="provider-card">
          <h3 class="provider-name">OpenRouter</h3>
          <span class="provider-url">https://openrouter.ai/api/v1</span>
          <span class="provider-key">Header: <code>Authorization: Bearer sk-or-...</code></span>
          <p class="provider-note">
            Use the <code>vendor/model</code> identifier
            (e.g. <code>meta-llama/llama-4-70b-instruct</code>).
            Reasoning-capable routes auto-detected.
          </p>
        </div>
        <div class="provider-card">
          <h3 class="provider-name">Together AI</h3>
          <span class="provider-url">https://api.together.xyz/v1</span>
          <span class="provider-key">Header: <code>Authorization: Bearer ...</code></span>
          <p class="provider-note">
            OpenAI-compatible. Good for open-weight models.
          </p>
        </div>
        <div class="provider-card">
          <h3 class="provider-name">Fireworks AI</h3>
          <span class="provider-url">https://api.fireworks.ai/inference/v1</span>
          <span class="provider-key">Header: <code>Authorization: Bearer ...</code></span>
          <p class="provider-note">
            OpenAI-compatible. Fast open-weight inference.
          </p>
        </div>
        <div class="provider-card">
          <h3 class="provider-name">Groq</h3>
          <span class="provider-url">https://api.groq.com/openai/v1</span>
          <span class="provider-key">Header: <code>Authorization: Bearer gsk_...</code></span>
          <p class="provider-note">
            OpenAI-compatible. Low-latency inference.
          </p>
        </div>
        <div class="provider-card provider-card-shim">
          <h3 class="provider-name">Custom / self-hosted</h3>
          <span class="provider-url">Any OpenAI-compatible URL</span>
          <span class="provider-key">Header: <code>Authorization: Bearer ...</code> (or none)</span>
          <p class="provider-note">
            vLLM, SGLang, Ollama (OpenAI adapter), TGI, llama.cpp HTTP
            server, etc. Provide a publicly reachable base URL.
          </p>
        </div>
      </div>
      <p class="provider-footnote">
        Missing your provider? Pick <code>custom</code> and paste the
        base URL, or reach out at
        <a href="mailto:contact@terms-bench.org">contact@terms-bench.org</a>.
      </p>
    </section>

    <!-- ===================================================== Notes -->
    <h2 class="submit-section-h">Fine print</h2>
    <ul class="submit-notes-list">
      <li>
        <strong>Cost.</strong> The full paper suite is roughly
        <code>3 &times; 6 &times; 60</code> episodes &times; ~12 rounds &times;
        ~600 tokens/round &asymp; <code>~7-8M tokens</code> total. Frontier
        chat models land in the <code>$50-$120</code> range; reasoning
        models with high effort can reach <code>$200+</code>.
      </li>
      <li>
        <strong>Cadence.</strong> We batch submissions weekly. Expedited
        runs (e.g. for paper deadlines) can be arranged - email
        <a href="mailto:submissions@terms-bench.org">submissions@terms-bench.org</a>.
      </li>
      <li>
        <strong>Re-runs.</strong> Each model gets one re-run per six
        months (e.g. after a checkpoint update). Re-submit using the
        same display name to overwrite; pick a new display name to keep
        both versions on the leaderboard.
      </li>
      <li>
        <strong>Embargo.</strong> If you need results held until a
        publication deadline, mention it in the notes; we publish on
        your timeline up to 60 days.
      </li>
      <li>
        <strong>Reproducibility.</strong> Submissions ship with the
        exact <code>termsbench/defaults.toml</code> hash and the run
        manifest; everything is reproducible from the artifacts.
      </li>
      <li>
        <strong>Get in touch.</strong> General queries, interest,
        collaboration, or anything else - email
        <a href="mailto:contact@terms-bench.org">contact@terms-bench.org</a>
        and we'll do our best to get back.
      </li>
    </ul>

  </main>

  <script>
    // Theme toggle: matches index.html behavior.
    (function () {
      var btn = document.getElementById('theme-toggle');
      if (!btn) return;
      btn.addEventListener('click', function () {
        var current = document.documentElement.getAttribute('data-theme') || 'light';
        var next = current === 'dark' ? 'light' : 'dark';
        document.documentElement.setAttribute('data-theme', next);
        try { localStorage.setItem('terms-theme', next); } catch (_) {}
      });
    })();

    // Provider -> default base URL.
    var PROVIDER_BASE_URLS = {
      openai:     'https://api.openai.com/v1',
      anthropic:  'https://api.anthropic.com/v1',
      openrouter: 'https://openrouter.ai/api/v1',
      together:   'https://api.together.xyz/v1',
      fireworks:  'https://api.fireworks.ai/inference/v1',
      groq:       'https://api.groq.com/openai/v1',
      custom:     ''
    };

    var providerEl = document.getElementById('f-provider');
    var baseUrlEl  = document.getElementById('f-base-url');
    providerEl.addEventListener('change', function () {
      var def = PROVIDER_BASE_URLS[providerEl.value] || '';
      // Only autofill if the user hasn't typed something custom.
      if (!baseUrlEl.value || Object.values(PROVIDER_BASE_URLS).indexOf(baseUrlEl.value) !== -1) {
        baseUrlEl.value = def;
      }
    });

    // The mailbox that accepts submissions.
    var SUBMIT_EMAIL = 'submissions@terms-bench.org';

    var form     = document.getElementById('submit-form');
    var btn      = document.getElementById('submit-btn');
    var statusEl = document.getElementById('submit-status');

    function setStatus(state, msg) {
      statusEl.textContent = msg || '';
      if (state) statusEl.setAttribute('data-state', state);
      else statusEl.removeAttribute('data-state');
    }

    function val(id) { return document.getElementById(id).value.trim(); }

    form.addEventListener('submit', function (ev) {
      ev.preventDefault();
      if (!form.checkValidity()) {
        setStatus('err', 'Please fill in all required fields.');
        form.reportValidity();
        return;
      }
      var displayName = val('f-display');
      var subject = 'TERMS-Bench submission: ' + displayName;
      var lines = [
        'TERMS-Bench submission',
        '======================',
        '',
        'Submitter:    ' + val('f-name'),
        'Email:        ' + val('f-email'),
        'Affiliation:  ' + (val('f-affiliation') || '(none)'),
        '',
        'Display name: ' + displayName,
        'Model id:     ' + val('f-model'),
        'Provider:     ' + val('f-provider'),
        'Base URL:     ' + (val('f-base-url') || '(provider default)'),
        '',
        'Notes / API-key delivery / sampling overrides:',
        '----------------------------------------------',
        val('f-notes'),
        '',
        '-- ',
        'I authorize TERMS-Bench to run the full paper suite using my ' +
        'API access and to publish aggregate results on the public ' +
        'leaderboard, with a private report and a leaderboard PR for ' +
        'sign-off before publication.'
      ];
      var body = lines.join('\n');
      var href = 'mailto:' + SUBMIT_EMAIL +
                 '?subject=' + encodeURIComponent(subject) +
                 '&body=' + encodeURIComponent(body);

      // Hand off to the user's mail client.
      window.location.href = href;
      setStatus('ok',
        'Email composed in your mail client. Review and send to ' + SUBMIT_EMAIL + '.');
    });
  </script>
</body>
</html>