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 · pageplvw0217_29.png

What is the page number?

OCR text of the page · 847 characters; the scanned image itself is not published
29
Table 3. Possible Reasons for the Discrepancy Between the Findings of Observational Studies
and Those of Randomized Trials in the Prevention of CHD in Postmenopausal Women
HERS/ERA
Observational
CHD status
Secondary prevention of
Primary prevention of CHD
established CHD
Age of participants
Mean 265 years
Mean = ~55-60 years
Treatment
CEE plus progestins
Usually unopposed estrogens
Study design/Selection bias
Random assignment to HRT
Self-selection bias--HRT
or placebo
users younger, healthier
Event-related dropout
Excess early CV events
Possible loss of those with
early CV events
Observation time
Shorter
Longer
Time since menopause
Mean = 23.1 years
Mean <5 years
Method of analysis
Intent-to-treat
On-treatment
Confidential Pursuant to Confidentiality
SOLOJO10-002105
Order
Source: https://www.industrydocuments.ucsf.edu/docs/plvw0217
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
29
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

29

2 characters3 tokens

Aux 2015

Exact match

29

2 characters3 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

OCTOBER, 2000

Question:

What is the page number?

OCTOBER, 2000

Question:

What is the page number?

OCTOBER, 2000

Question:

What is the page

157 characters64 tokens