Claim · 2018 · event · Stated in source · Supported

Radford et al. (2018) describe generative pre-training of a language model on unlabeled text followed by discriminative fine-tuning on each task.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

Improving Language Understanding by Generative Pre-Training → discusses →

Where it appears

No collection page holds this claim. It is listed with its records in the explorer.

Claim and evidence

Improving Language Understanding by Generative Pre-Trainingdiscusses ↓

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source

Date: 2018 · event Records dated 2018 →

Scope and limitsA method statement from the abstract; the paper does not call its model a large language model.

The evidence 01

primary sourceSupports the claim

Improving Language Understanding by Generative Pre-Training

OpenAI · Publication date unknown

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative fine-tuning on each specific task.”

Location in source PDF p. 1, abstract

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-31

Source details and check notes

Authors: Alec Radford; Karthik Narasimhan
Version: Technical paper PDF served by OpenAI; the PDF prints no date. OpenAI’s release page of 11 June 2018 links it (cross-family check, item A005). Retrieved for the candidate pack on 2026-10-28 (SHA-256 eb81a65e0856bd38e855e1f1de6ee12e7c3eb92ae05c8270d345a475beb631d0). The author list is the byline as printed on the first page.
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: Alec Radford, Karthik Narasimhan, “Improving Language Understanding by Generative Pre-Training”.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, Kinds of AI and technical history (research/technical-history). The quotation (30 words) matched the text retrieved for the candidate pack on 2026-10-29. Matching ignored differences of whitespace, case, quotation marks, dashes, PDF line-break hyphens and Unicode compatibility forms, so the source’s typography can differ. No human editorial review.

Status and meaning

Editorial status. Not reviewed by an editor. An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.

Relation. The concept is explicitly discussed in the document. This makes no earliest-use, invention or transmission claim.

Date. Date as a later account or citation gives it; no contemporaneous text read here dates it. The PDF is undated; the year is the one later citations give.

Record ID th-fam-tllm-gpt1-def