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

From the 100, how many of them had baseline angiogram?→ 81 / N=81 How many subjects were enrolled?→ 100 / N=100 Why were 3 rejected?→ Protocol violation / Protocol Violation How many Subjects signed out the consent form?→ 103 / N=103 what is the effective date ?→ 3-19-01 who is the RJRT authorised agent?→ Terry Whitson / terry whitson to whom terry whitson is delivering ?→ Phipps Bend who is the grower?→ Ronnie Hurd / ronnie hurd what is the name of the company?→ R. J. Reynolds Tobacco Company who signed the grower signature?→ Ronnie Hurd / ronnie hurd what is the zip code ?→ 37731 what is the effective date?→ 3-19-01 what is the delivery point?→ Phipps Bend

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagenpwx0225_10.png

to whom terry whitson is delivering ?

OCR text of the page · 318 characters; the scanned image itself is not published

Attachment 2
Effective Date
31901
RJRT's Authorized Agent and Point of Delivery
1.
RJRT Authorized Agent:
Terry Whitson
2.
Point of Delivery:
Phipps Bend
110 Rae have
burgoins ville, In 32873
Grower:
Ronnie Blud
RJRT Representative:
52381 9369
Page 10 of 12
Source: https://www.industrydocuments.ucsf.edu/docs/npwx0225

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

Phipps Bend

Ours

Wrong

Grower: Ronnie Blud

Aux 2015

Wrong

Terry Whitson

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

RJRT Representative:

Input:

RJRT Representative:

Input:

RJRT Representative:

Input:

RJRT Representative:

Output Explorer

Ours

Qwen 3.5 27B

Aux 2015

PiT-FT 2015

ChronoGPT 2015

Input:

Input:

Input:

Response