Financial NLP
Financial News Multisource
News at volume for monitoring, language modelling, and signal extraction rather than document QA.
As of 13 August 2026. Conditional pack. Gated multi-source financial news corpus, 1990 to 2025. Not a document-QA pack.
This corpus is for news at volume: monitoring, language modelling, and signal research. It is not TAT-QA and not a 10-K reasoning set. Take a sample subset; do not treat 57 million rows as one accuracy.
Gated Hugging Face terms must be accepted. No Fable 5 or GPT-5.6 table is cited.
- Scale
- Multi-source financial news spanning 1990 to 2025. Pin a sample subset at acquisition.
- Provenance
- Gated Hugging Face financial-news corpus. Take a sample subset; do not treat the full dump as a QA pack.
- Contamination
- High. Long public news history.
- Cannot establish
- Alpha, a research-terminal score, or a full-corpus commercial run before gated terms are accepted.
Metrics, kept separate
- Subset-specific official metrics
- Sampled classification accuracy
- Coverage by source
No cited public figure is installed for this pack. Method, scale, and gates still apply. A missing chart is not a pass rate.
What the figures bound
The pack is listed as a volume test for news systems. Volume is not QA accuracy.
No chart is installed. A sampled subset would need its own frozen protocol.
What this page does not conclude
- A commercial full-corpus run before gated terms are accepted
- Trading-signal or alpha quality
- A research-terminal score
- GPT-5.6 or Fable 5 news-at-volume performance
Cited as calibration on these memos — not a score of those products.
Not attached to a software, advisory, or automation memo yet.