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 · pagegscv0228_6.png

Who is leading the discussion on 'avenues for research' ?

OCR text of the page · 870 characters; the scanned image itself is not published
THURSDAY 22ND SEPTEMBER
"EFFECTS ON HEALTH OF ALCOHOL CONSUMPTION"
Chairman: Professor W Ironside
9.00
"The Borderline Between Safe and Hazardous Drinking"
Dr Griffiths Edwards
Discussion
10.30 - 10.45
COFFEE
10.45
"Social Class in Epidemiology - A Need for Reclassification"
Professor Sir Cyril Clarke
Discussion
11 . 30
"The Australian Study of Drinking Patterns"
Professor R Blackett
Discussion
12.30
"Avenues for Research"
Discussion led by Professor T Turner
13.00 - 14.00
LUNCH
"FOETAL DAMAGE 'AND "ALCOHOL CONSUMPTION"
Chairman: Professor G Heseltine
14.00
"A Review"
Professor W Ironside
Discussion
15.00
"Nutritional Aspects"
Professor W Darby
Discussion
"BEER COMPONENTS"
15. 40
"Congeners - do they matter?"
Professor W D Paton
Discussion
19. 15 for 19.45-
DINNER AT THE PARK ROYAL BREWERY OF ARTHUR GUINNESS SON & COMPANY LIMITE
19.30 FOR 20.00
Source.
SCVO2
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
professor t turnerProfessor T Turner
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.50

Discussion led by Professor T Turner

36 characters7 tokens

Aux 2015

Exact match

Professor T Turner

18 characters4 tokens

PiT-FT 2015

Wrong

<|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|>

120 characters64 tokens

ChronoGPT 2015

Wrong

ChronoGPT

Question:

Who is leading the discussion on "avenues for research" ?

ChronoGPT

94 characters35 tokens