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 · pagethcn0226_3.png

What is the 4th concern?

OCR text of the page · 285 characters; the scanned image itself is not published
. . . ......
...........
Cross-cutting concerns
1) Use of outdated and problematic Preamble
2) Criteria for evaluating study quality
3) Benchmark dose modeling transparency
4) Quantification of the Suggestive cancer endpoint
Source: https://www.industrydocuments.ucsf.edu/docs/thcn0226
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
quantification of the suggestive cancer endpointQuantification of suggestive cancer endpoint
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

Quantification of the Suggestive cancer endpoint

48 characters10 tokens

Aux 2015

Wrong

4th concern

11 characters4 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong
  1. Use of outdated and problematic Preamble
  2. Criteria for evaluating study quality
  3. Benchmark dose modeling transparency
124 characters27 tokens