Dataset
sec-filings-forward-return-2026
A leak-free, chronologically ordered corpus of 8-K, 10-Q, and 10-K filings for roughly 610 present-day U.S. large-cap companies, each paired with objective forward-return labels through 2026. The release is structured as a benchmark for testing whether filing language can anticipate subsequent stock performance, with clearly separated provenance classes.
Data preview
Dataset card
The sec-filings-forward-return-2026 dataset, published on the Hugging Face Hub by user BatuhanECB, pairs leak-free SEC 8-K, 10-Q and 10-K filings for roughly 610 current US large-cap names with objective forward-return labels. It is structured as a benchmark for testing whether filing text can predict post-filing stock returns.
What this dataset contains
The dataset bundles time-ordered SEC filings for approximately 610 US large-cap issuers spanning 2000 through 2026, together with market-relative forward-return labels at horizons of 1, 5, 20, 60 and 120 trading days. Labels are provided as excess returns versus SPY (`excess_spy`), and the corpus-only configuration is released under CC0. Splits are walk-forward with a 120-trading-day embargo: train covers up to 2019, validation covers 2020 to 2022, and test covers 2023 to 2026, with any filing whose label window crosses a split boundary dropped from the earlier split.
Structure and fields
Each row corresponds to a single filing identified by `accession`, with the issuer represented by `cik` and `ticker`, the filing type stored in `form`, and the SEC filing date in `file_date`. The raw filing text is held in `text`, with its character count recorded in `text_len`; trading-day-aware timing fields including `entry_date` are retained for transparency. The dataset ships in Parquet format with mixed tabular and text modalities, totalling 631,426 rows across the train, validation and test splits.
How analysts use it
Quantitative researchers can use the dataset to benchmark text-classification and tabular-regression models that attempt to extract alpha from filing language at specific horizons, with the 8-K short-horizon setting flagged as the most promising starting point. Long/short equity managers may also evaluate whether event-driven signals embedded in 10-K and 10-Q text add value to a SPY-relative return forecast, subject to the baseline evidence that out-of-sample walk-forward AUC is close to 0.50.
Documentation & references 1
Related datasets
Browse moreParsed eligibility criteria from 13,229 ClinicalTrials.gov studies representing the candidate pool gathered by two first-stage retrievers during TREC Clinical Trials 2021–2023, formatted as typed entity-relation graphs for reranking models.
The clinical-trials-patient-graphs dataset is a small tabular collection of patient-level clinical records distributed in Parquet format by the 2001jdev publisher. It organizes entities and relations extracted from trial data into structured rows spanning 2021 through 2023.
This dataset is a parsed snapshot of ClinicalTrials.gov records, distributed in Parquet format on the Hugging Face Hub under an unspecified license. It contains roughly 52,000 trial entries drawn from public registry filings.
clinical-trials-trec-qrels is a tabular relevance-judgment file distributed by the 2001jdev user on the Hugging Face Hub. It maps clinical-trial topic identifiers to NCT registry IDs with graded relevance scores.
The clinical-trials-trec-topics dataset is a small tabular collection of clinical-trial topic records distributed in Parquet format on the Hugging Face Hub. It appears to be derived from TREC clinical-trial retrieval topics, with rows partitioned by year.
SEC EDGAR 10-K & 10-Q financial statements for 2,600+ NASDAQ companies.
Discussion
Questions or observations about this dataset? Join the conversation on X.
No comments yet. Start the conversation on X.