data-pack
llama3_8b_learn_from_70b_data_n4_filter_2e6_bz32_pack8192_also_train_reward_2epochllama3_prompt_baseline_learn_from_70b_data_n4_filter_2e6_bz32_pack8192_3epochwo_packing_v2_1_pt_ep1_sft_ep1_based_on_ourmodel_with_all_choosen_inst_data_20240725llama3_8b_learn_from_70b_data_n4_filter_2e6_bz32_pack8192_also_train_reward_3epochwo_packing_v2_1_pt_ep1_sft_ep5_based_on_ourmodel_with_all_choosen_inst_data_20240725llama3_8b_self_gen_data_n40_filter_2e6_bz32_pack8192_3epochwo_packing_v2_1_pt_ep1_sft_ep5_based_on_mergekit_with_all_choosen_inst_data_20240726wo_packing_v2_1_pt_ep1_sft_ep3_based_on_mergekit_with_all_choosen_inst_data_20240726
arb-officeqa-data-packet
ARB OfficeQA data packet
The training data packet of Applied RSI Bench's (ARB) OfficeQA benchmark: the documents a
post-training agent may study before it trains a model for OfficeQA, Databricks' questions on
the U.S. Treasury Bulletin (1939-2025). One archive, L.tar.gz, built by ARB's data_packet
pipeline (map, scout, script, list, build) from OfficeQA's questions without their answers,
kept to sources whose terms allow redistribution, plus OfficeQA's own corpus.… See the full description on the dataset page: https://huggingface.co/datasets/junlinw/arb-officeqa-data-packet.realigns-business-open-data-pack
Realigns Business Open Data Pack
A developer-friendly open business dataset pack prepared by Realigns Inc. for AI assistants, dashboards, market research tools, RAG systems, financial research tools, HR analytics, and business intelligence applications.
This repository combines multiple public business and economic datasets into clean CSV and JSONL formats.
Dataset Index
For a complete folder-by-folder overview of all included datasets, see:
DATASETS.md… See the full description on the dataset page: https://huggingface.co/datasets/realigns/realigns-business-open-data-pack.Packora-data
Packora data
Project links
Project page
Try Packora (live demo)
Paper
Code
Model checkpoints
Hugging Face collection
This repository contains the CSV refcode manifests used to reconstruct the
Packora training and benchmark datasets with a licensed Cambridge Structural
Database installation. It also provides the aggregate dataset statistics used
by the released models. It does not contain CSD structures or prebuilt LMDBs.
csv_manifests/csd.csv is the Packora… See the full description on the dataset page: https://huggingface.co/datasets/nayoung10/Packora-data.arb-fav2-data-packet
fav2 data packet
Study material for the fav2 benchmark of Applied RSI Bench (ARB): the documents an analyst would read to answer the 27 public Vals AI Finance Agent v2 questions (SEC filings of the companies the questions name and of the wider US market, federal agency pages and data, regulations), as
plain text where possible. An ARB post-training agent gets the packet read-only at
/home/agent/workspace/data_packet/ before its model is evaluated.
This is a redistributable… See the full description on the dataset page: https://huggingface.co/datasets/junlinw/arb-fav2-data-packet.skirmish-datapacks
Skirmish data packs
Private data packs consumed by the Skirmish
continuous-integration system tests. They are produced by the
gameplay_export tool in cornerian/skirmish-assets from a verified
Super Smash Bros. Melee (USA 1.02) disc extraction, using the layouts of the
pinned doldecomp/melee revision named in each pack's manifest.json.
Path
Contents
gameplay/v1/skirmish-gameplay-v1.tar.gz
Minimal headless gameplay data: Fox (attributes, bones, hurtboxes, ECB, jab) and… See the full description on the dataset page: https://huggingface.co/datasets/cornerian/skirmish-datapacks.arb-harvey-lab-data-packet
Harvey LAB data packet
Study material for the harvey-lab benchmark of Applied RSI Bench (ARB): the
documents a lawyer would read to do the work in
Harvey LAB's tasks (statutes,
regulations, agency guidance, court opinions, filings, model documents and
practice material), as plain text where possible. An ARB post-training agent gets
the packet read-only at /home/agent/workspace/data_packet/ before its model is
evaluated.
The packet holds no LAB scenarios, case files, task… See the full description on the dataset page: https://huggingface.co/datasets/junlinw/arb-harvey-lab-data-packet.