Spaces:
Running
Running
Add FlavourBench co-authors
Browse filesAdd the canonical paper byline to the Space card, hero, and footer.
README.md
CHANGED
|
@@ -14,6 +14,8 @@ datasets:
|
|
| 14 |
|
| 15 |
# FlavourBench
|
| 16 |
|
|
|
|
|
|
|
| 17 |
An interactive explorer for the powered FlavourBench release: 20 frontier endpoints, 640
|
| 18 |
executable culinary decisions, 12,800 primary responses, 1,280 label-permuted repeats, and all 190
|
| 19 |
paired statistical comparisons.
|
|
@@ -32,3 +34,13 @@ The Space reads one content-addressed bundle and makes no provider calls. Use it
|
|
| 32 |
The dataset contains full response records. The Space carries compact excerpts for fast browsing.
|
| 33 |
See the repository [rights boundary](https://github.com/josefchen/flavourbench/blob/main/LICENSES.md)
|
| 34 |
for component-level licensing.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
# FlavourBench
|
| 16 |
|
| 17 |
+
**Josef Chen 路 Jakub Radzikowski 路 Erim Hayretci**
|
| 18 |
+
|
| 19 |
An interactive explorer for the powered FlavourBench release: 20 frontier endpoints, 640
|
| 20 |
executable culinary decisions, 12,800 primary responses, 1,280 label-permuted repeats, and all 190
|
| 21 |
paired statistical comparisons.
|
|
|
|
| 34 |
The dataset contains full response records. The Space carries compact excerpts for fast browsing.
|
| 35 |
See the repository [rights boundary](https://github.com/josefchen/flavourbench/blob/main/LICENSES.md)
|
| 36 |
for component-level licensing.
|
| 37 |
+
|
| 38 |
+
## Citation
|
| 39 |
+
|
| 40 |
+
```bibtex
|
| 41 |
+
@article{chen2026flavourbench,
|
| 42 |
+
title = {FlavourBench: Ranking Frontier Language Models with Executable Culinary Ground Truth},
|
| 43 |
+
author = {Chen, Josef and Radzikowski, Jakub and Hayretci, Erim},
|
| 44 |
+
year = {2026}
|
| 45 |
+
}
|
| 46 |
+
```
|
app.py
CHANGED
|
@@ -78,6 +78,13 @@ body, .gradio-container {
|
|
| 78 |
margin: 0;
|
| 79 |
max-width: 640px;
|
| 80 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
.fb-stats {
|
| 82 |
display: grid;
|
| 83 |
grid-template-columns: repeat(4, minmax(92px, 1fr));
|
|
@@ -345,6 +352,7 @@ def _hero_html() -> str:
|
|
| 345 |
<h1>640 decisions.<br>No model judge.</h1>
|
| 346 |
<p class="fb-dek">Executable score maps rank 20 frontier endpoints with shared-task
|
| 347 |
uncertainty and inspectable responses.</p>
|
|
|
|
| 348 |
<div class="fb-stats">
|
| 349 |
<div class="fb-stat"><strong>{len(MODELS)}</strong><span>models</span></div>
|
| 350 |
<div class="fb-stat"><strong>{len(TASKS)}</strong><span>tasks</span></div>
|
|
@@ -689,7 +697,7 @@ make -C paper -f Makefile.powered arxiv
|
|
| 689 |
gr.HTML(
|
| 690 |
"""
|
| 691 |
<div class="fb-shell fb-footer">
|
| 692 |
-
FlavourBench |
|
| 693 |
</div>
|
| 694 |
"""
|
| 695 |
)
|
|
|
|
| 78 |
margin: 0;
|
| 79 |
max-width: 640px;
|
| 80 |
}
|
| 81 |
+
.fb-byline {
|
| 82 |
+
color: var(--fb-ink);
|
| 83 |
+
font-size: 14px;
|
| 84 |
+
font-weight: 650;
|
| 85 |
+
letter-spacing: .01em;
|
| 86 |
+
margin: 16px 0 0;
|
| 87 |
+
}
|
| 88 |
.fb-stats {
|
| 89 |
display: grid;
|
| 90 |
grid-template-columns: repeat(4, minmax(92px, 1fr));
|
|
|
|
| 352 |
<h1>640 decisions.<br>No model judge.</h1>
|
| 353 |
<p class="fb-dek">Executable score maps rank 20 frontier endpoints with shared-task
|
| 354 |
uncertainty and inspectable responses.</p>
|
| 355 |
+
<p class="fb-byline">Josef Chen 路 Jakub Radzikowski 路 Erim Hayretci</p>
|
| 356 |
<div class="fb-stats">
|
| 357 |
<div class="fb-stat"><strong>{len(MODELS)}</strong><span>models</span></div>
|
| 358 |
<div class="fb-stat"><strong>{len(TASKS)}</strong><span>tasks</span></div>
|
|
|
|
| 697 |
gr.HTML(
|
| 698 |
"""
|
| 699 |
<div class="fb-shell fb-footer">
|
| 700 |
+
FlavourBench | Josef Chen 路 Jakub Radzikowski 路 Erim Hayretci
|
| 701 |
</div>
|
| 702 |
"""
|
| 703 |
)
|