Output Explorer

Every prompt in the paper, and what each model wrote back.

Answer a question about a scanned page from its OCR text. Scored by exact match and ANLS against the accepted answers.

13 of 5,330 prompts

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagekzng0227_22.png

What is the no. of samples for Rockfish (R)?

OCR text of the page · 780 characters; the scanned image itself is not published
-20-
Sodium
Potassium
mg. / 100g.
No of Refer-
FOOD
Max.
Min.
AV.
mg./100g. . Samples
ence
FISH AND SHELLFISH, con't.
Clams (R)
603
172
5
Cod, raw (R)
60
360
1
Cod, frozen fillets (0)
400
LOO
1
Cod, liver oil (R)
0.1
Cod, salted, dried (0)
8100
160
Codfish (R)
75
339
5
Crab, boiled in fresh water (0)
366
271
6
Crab, canned (0)
1000
1.10
Eels, various (R)
267
237
6
Flounder (R)
68
31.1.
5
Haddock (R)
125
302
6
Halibut, raw (R)
56
340
1
Herring (R)
100
303
5
Herring, fillets (R)
130
317
6
Lobster, boiled in fresh water (0)
325
258
6
Lobster, boiled in tap water (0)
210
180
Mussels (R)
289
315
6
Oysters (R)
73
110
Oysters (R)
73
Oysters (R)
505
258
6
Rockfish (R)
45
334
2
Salmon (R)
53
331
5
Salmon, raw (R)
48
410
1
Salmon, canned (0)
540
300
1
Source: https://www.industry
System prompt · identical for every setup
Answer the question using only the OCR text from a single document page. Return only the answer, with no explanation. Preserve the answer wording from the OCR text when possible.
Expected answer
1
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

2

1 characters2 tokens

Aux 2015

Wrong

-20-

4 characters5 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER OCTOBER

103 characters64 tokens