REVIEW 2 cited by
Can Language Models Represent the Past without Anachronism?
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Before researchers can use language models to simulate the past, they need to understand the risk of anachronism. We find that prompting a contemporary model with examples of period prose does not produce output consistent with period style. Fine-tuning produces results that are stylistically convincing enough to fool an automated judge, but human evaluators can still distinguish fine-tuned model outputs from authentic historical text. We tentatively conclude that pretraining on period prose may be required in order to reliably simulate historical perspectives for social research.
Forward citations
Cited by 2 Pith papers
-
Where Models Converge and Humans Diverge: A Coverage Framework for Distributional Pluralism in Open-Ended Generation
LLM outputs are usually plausible but cover far less of the human response space than matched human samples do, with the biggest gap at the periphery of the human distribution.
-
TimeCapsule: Generative Hallucination as a Method for Historical Sensemaking
Training a language model exclusively on pre-1875 texts is proposed as a way to make its anachronistic errors readable as Victorian sensemaking.
Discussion (0). Continue with ORCID to comment.