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 · pageptkg0227_32.png

What is the body weight of dog named "11M" in grams?

OCR text of the page · 673 characters; the scanned image itself is not published
32
D-4-c. Data on Histopathology: organ weights for all animals
sacrificed.
Weights of Organs (gm. ) at Necropsy
Dog
Treat-
Kidney
Testes
ment
Body Wt. Heart Liver Spleen
Right Left
Right
Left
F*
BO
12028
112
335
102
60
(Combined wt. )
27F
BO
9534
87
408
77
31
31
6M
BO
12995
107
325
65
32
32
11
11
8M
BO
8850
78
246
69
26
26
16
16
3F
B3
10582
96
370
98
33
33
25F
B3
10328
87
388
49
30
30
11M
B3
8909
71
231
49
23
23
9
9
12M
B3
10268
90
265
79
21
20
10
10
5F
B6
10215
74
266
42
21
21
26F
B6
14613
109
485
43
33
32
7M
B6
7774
72
216
61
20
19
10
10
13M
B6
11009
105
317
62
35
35
13
14
* F = Female
M = Male
re - 35
Source: https://www.industrydocuments.ucsf.edu/docs/ptkg0227
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
8909
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

8909

4 characters5 tokens

Aux 2015

Wrong

We are given the question: "What is the body weight of dog named '11M' in grams?"

We are given the OCR text:

"32 D-4-c. Data on Histopathology: organ weights for all animals sacrificed. Weights of Organs (

207 characters64 tokens

PiT-FT 2015

Wrong

<|_assistant|> <|_assistant|> <|_assistant|> <|_assistant|> <|_assistant|> <|_assistant|> <|_assistant|> <|_assistant|>

120 characters64 tokens

ChronoGPT 2015

Wrong

Dog Body Weight (gm. ) at Necropsy Organ Weight (gm. ) at Necropsy Right Left Left Left Right Left Left Right Body Wt. Heart Liver Spleen Right Right Left Left Left Left Left Left

179 characters64 tokens