krummholzAI

The README is part
of the dataset.

Source terms. Cleaning notes. Checksums. Known limits. Explore preliminary datasets for forestry, fishing, customer service, robotics and LLM pipelines, with the documentation kept alongside the downloads.

Preliminary Catalog updated October 26, 2026 5 focus areas · source-specific releases

These are research starting points, with automated cleaning and documented limitations. They have not completed independent domain review or been validated for production deployment.

Forest Service ecologist Sjana Schanning measuring a tree diameter in snow-covered woods.
USDA FIA fieldwork in Wisconsin. USDA Forest Service photo by Ben Nurre · License.

Start with a free sample.

The banking and forestry samples remain free, with their source terms, documentation and limits. Larger curated forestry, fishing and robotics packages are $14.99 each, plus applicable sales tax.

Free banking sampleFree forestry samplePurchase and reuse terms
01

Forestry

Forest inventory features for exploring species, size and condition.

Public forest inventory Preliminary

Maine FIA forest measurements

Maine forest inventory measurements for exploratory models of tree dimensions, biomass, carbon and forest condition. The two tree source parts are combined before deduplication. All available years are retained.

903,254 unique feature rows
Format
CSV feature tables + species codebook
License
USDA public-use data - CC0 identified in USDA’s dataset record
Source
USDA Forest Service FIA

Represents 1,095,255 tree and condition observations, with repeated feature rows stored as record_count. Download both parts. Multi-year sample; precise locations and record linkages removed. Not a current tree census or a basis for statewide estimates.

Buy dataset · $14.99

$14.99 plus applicable sales tax. Includes both ZIP parts. Documentation and source licenses remain public.

Read documentation Field dictionary Cleaning audit
02

Fishing

Recreational fisheries observations for exploratory classification and modeling.

Recreational fishing survey Preliminary

NOAA MRIP 2024 fishing observations

NOAA Marine Recreational Information Program 2024 recreational fishing observations across all six available waves. Catch, trip activity and measured fish size support exploratory tabular ML. These are recreational survey data, not commercial fishing or vessel-control training data.

295,104 unique feature rows
Format
CSV feature tables + species codebook
License
NOAA U.S. government public-use data
Source
NOAA MRIP

Represents 392,082 retained catch, trip and size observations. Repeated feature rows use record_count. Broad state/area fields remain; fine locations and survey linkage keys are removed. Does not reproduce NOAA population estimates.

Buy dataset · $14.99

$14.99 plus applicable sales tax. Includes the complete curated ZIP. Documentation and source licenses remain public.

Read documentation Field dictionary Cleaning audit
03

Customer service

Intent examples for banking support, purchases, payments, delivery, complaints and refunds.

Labeled intent queries Preliminary

BANKING77 banking customer-service intents

Banking customer queries across all 77 intents, including card support, payments, transfers, cash withdrawals, fees, identity verification and refunds.

13,006 queries · 77 intents
Format
CSV or JSONL (same records)
License
CC-BY-4.0
Source
Publisher dataset · License terms

Automated privacy screening changes some wording. Queries and intent labels only; source evaluation splits are retained but modified. No guarantee of complete anonymization or removal of semantic paraphrases.

Download ZIP · 848 KB
Read documentation Cleaning audit
Labeled intent queries Preliminary

CLINC150 shopping and payment intents

Crowdsourced requests for order placement, order status and declined cards. The other intents and out-of-scope examples are excluded.

447 queries · 3 intents
Format
CSV or JSONL (same records)
License
CC-BY-3.0
Source
Publisher dataset · License terms

Automated privacy screening changes some wording. Queries and intent labels only; source evaluation splits are retained but modified. No guarantee of complete anonymization or removal of semantic paraphrases.

Download ZIP · 38 KB
Read documentation Cleaning audit
04

Workplace robotics

Numeric manipulation trajectories for research on robot learning.

Numeric robot demonstrations Preliminary

Jaco Play - numeric trajectories

1,085 complete tabletop-robot episodes with 77,965 ordered numeric frames and 89 task instructions, extracted from the pinned LeRobot conversion. Images and video excluded.

1,085 episodes · 77,965 frames
Format
CSV frames, episodes and task instructions
License
CC BY 4.0
Source
CLVR Jaco Play · LeRobot conversion

Full numeric inventory of the pinned LeRobot Jaco conversion. Images and video excluded. Coordinate units are not independently verified; no independent evaluation split. Offline research data, not validated real-robot control commands.

Buy dataset · $14.99

$14.99 plus applicable sales tax. Includes numeric frames, episodes and task instructions. Documentation and source licenses remain public.

Read documentation Field dictionary Cleaning audit
05

LLM training

Inspect the source text before you scale a language-model pipeline.

Historical text · Common Corpus Preliminary

Historical Reasoning & Forestry Pilot

A small, inspected selection of historical prose: 30 excerpts about reasoning and language, plus seven about forestry and botany. Use it to explore ingestion, provenance and retrieval before developing a larger training collection.

37 curated excerpts
8,801 words, approximately
3 source fragments
Format
UTF-8 JSONL + schema, source register and audit
Source works
Two books and one journal volume; selected excerpts only
Reuse basis
Documented U.S. public-domain assessment; jurisdiction-specific limits apply

A curation pilot, not a representative training corpus or a validated benchmark. No model has been trained or evaluated on this release. Historical claims and residual OCR errors may remain.

Download complete ZIP · 47 KB Text records · JSONL
Documentation Field dictionary Sources & rights evidence Cleaning audit Reuse notes Sizes & SHA-256 checksums

Explore the LLM training area: methods, sample text and reuse limits →

Privacy revision · October 23, 2026. Banking downloads and the older homepage samples have been updated. Read the independent review and changes →

06

Synthetic Data

Original generated test records, clearly separated from real observations and upstream collections.

Krummholz-generated · Fully synthetic Preliminary

Maine & northern New England forestry test data

Fictional forestry tables and notes for data-ingestion, cleaning and validation experiments, released alongside the Python generator under MIT.

Generated dataset ZIP + Python script · MIT

Includes intentionally malformed records and inconsistent values. Not independently field- or domain-validated; not for production forestry prescriptions. The defect manifest is not exhaustive. This original release is distinct from curated upstream datasets.

Downloads, documentation & limitations →

Bespoke datasets from your existing data.

Projects are priced above the $14.99 catalog releases and quoted after the task, data permissions, cleaning, annotation and delivery requirements are agreed.

Discuss a bespoke dataset

Open data.
Specific terms.

Each dataset keeps its documented upstream terms. MIT, Creative Commons, public-domain and other open-data terms are not interchangeable. Cleaning or annotating a source does not remove its attribution or redistribution requirements.

These dataset terms are separate from Bristlecone’s MIT software license. Consult the included license, provenance and modification notices before reuse.

  • BANKING77 and Jaco Play numeric trajectories: Creative Commons Attribution 4.0 (CC BY 4.0).
  • CLINC150: Creative Commons Attribution 3.0 (CC BY 3.0).
  • USDA FIA forestry: CC0 identified in the USDA source record.
  • NOAA MRIP fishing: U.S. government public-use data under the documented source terms, not an MIT or CC license assigned by Krummholz.
  • Historical LLM-training pilot: documented U.S. public-domain assessment for source texts; CC0 for Krummholz additions to the extent rights can be granted.
  • Krummholz synthetic forestry test release and generator: MIT. Fully synthetic, preliminary and intentionally imperfect; not field observations.

No blanket MIT or Creative Commons license applies to the whole catalog. Each release’s own notices control; the summary above does not replace them.

Useful fields.
Traceable changes.

  • Task-focused records in open, machine-readable formats.
  • Documented duplicate handling and checks for identifying information.
  • Source attribution, license terms, field descriptions and cleaning counts.
  • Checksums so you can verify the files you downloaded.

Know the limits.

Automated screening does not certify anonymity or label accuracy. Review the included documentation before combining, splitting or training on a dataset. The support files omit answer targets; the ecological files omit precise locations and source linkage fields; the robotics release contains numeric trajectories without video.

Licenses differ by source. No single license applies to the entire catalog. Curated releases are derivatives of credited source datasets. Our separately labeled synthetic forestry release contains generated test records. Neither is an original Krummholz field collection.

Read the October 22 release report → Package sizes and SHA-256 checksums →