Scaling Point-in-Time Language Models

Bryan T. Kelly, Semyon Malamud, Johannes Schwab and Teng Andrea Xu

No 35247, NBER Working Papers from National Bureau of Economic Research, Inc

Abstract: Large language models trained on unrestricted internet corpora inevitably embed information from the future, introducing lookahead bias that compromises the validity of backtests and causal inference in finance and the social sciences. Point-in-time language models—trained exclusively on text available up to each calendar date—eliminate this leakage by construction, but existing efforts typically produce models that lag substantially behind their unconstrained counterparts. We show that this performance gap can be narrowed through scale. Training decoder-only transformers with up to 4 billion parameters on 1 trillion chronologically filtered tokens from FineWeb, we construct a sequence of monthly model checkpoints spanning 2013–2024. Across a range of common-sense reasoning and language understanding benchmarks, our models approach the performance of leading open-weight models of comparable size (such as Gemma-3-4B and LLaMA-7B) trained on temporally unrestricted data, although a performance gap remains on several tasks. Finally, in a strict out-of-sample economic evaluation task, portfolios built from point-in-time embeddings achieve robust positive Sharpe ratios and perform close to full-sample counterparts that violate temporal validity, indicating that chronologically consistent language models can extract economically meaningful signals without relying on look-ahead bias. We release the complete pipeline—including dataset construction, training infrastructure, and evaluation code—to enable reproducible point-in-time language modeling and to support research applications that require strict temporal validity.

JEL-codes: C14 C45 G11 G14 G17 (search for similar items in EconPapers)
Date: 2026-05
Note: AP
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.nber.org/papers/w35247.pdf (application/pdf)
Access to the full text is generally limited to series subscribers, however if the top level domain of the client browser is in a developing country or transition economy free access is provided. More information about subscriptions and free access is available at http://www.nber.org/wwphelp.html. Free access is also available to older working papers.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:nbr:nberwo:35247

Ordering information: This working paper can be ordered from
http://www.nber.org/papers/w35247
The price is Paper copy available by mail.

Access Statistics for this paper

More papers in NBER Working Papers from National Bureau of Economic Research, Inc National Bureau of Economic Research, 1050 Massachusetts Avenue Cambridge, MA 02138, U.S.A.. Contact information at EDIRC.
Bibliographic data for series maintained by ().