datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Openthoughts_math_30k_opsdopenthoughts_math_30k_opsd_splitopsd-cl-assets
opsd-cl-assets
Offline assets for the OPSD continual-learning project on a box without DNS.
Built 2026-09-12 on VAST. Pull everything with one command, then run install_on_box.sh.
path
content
size
wheelhouse/
167 wheels resolved from siyan-zhao/OPSD environment.yml for cp310 / manylinux x86_64, plus deepspeed-0.18.2.tar.gz and the lock file
4.7 GB
models/Qwen3-1.7B
Qwen/Qwen3-1.7B
3.8 GB
datasets/Openthoughts_math_30k_opsd
siyanzhao/Openthoughts_math_30k_opsd… See the full description on the dataset page: https://huggingface.co/datasets/HzChen20/opsd-cl-assets.Full_Agent_RL_OPSD_with_Just_2_A800sswe_opsd_datasetopsd-toolSPSD-Variants-opsd
SPSD-Variants-opsd
Grounded on-policy self-distillation (OPSD) teacher-context dataset over 45
board-game rule variants (5 families × 9: connect4, domineering,
simplified_first_attack, simplified_othello, tic_tac_chess), derived from
trained MuZero/EfficientZero checkpoints (plan-528 v2).
Each row is a decision-state task (a move choice or one of six auxiliary
state-QA tasks). The privileged_context is the teacher signal: grounded
natural-language reasoning that discovers the… See the full description on the dataset page: https://huggingface.co/datasets/LorMolf/SPSD-Variants-opsd.scaleswe-opsd-v2-3200-summary
Scale-SWE OPSD v2 — 3200 tasks with summary hints
The training set used for the Scale-SWE on-policy self-distillation (OPSD) runs. 3200 SWE tasks across
752 repositories, each paired with a reference agent trajectory and a condensed solution hint.
Uploaded from /checkpoint/huggingface/datasets/scaleswe_opsd_v2_3200_summary (a
datasets.save_to_disk directory), converted to parquet. Row count, ids and field contents verified
identical to the source.
⚠️ Contains… See the full description on the dataset page: https://huggingface.co/datasets/starli-snowflake/scaleswe-opsd-v2-3200-summary.opsdataDataset Card for Energy — OPEN POWER System Data
This dataset was prepared for the OpenHPI course Time Series Analysis and Forecasting and provides electricity consumption, renewable generation, weather, and electricity price data across European countries, including Germany. The data is aggregated from the ENTSO-E Transparency Platform and the Open Power System Data project, spanning over 10 years with 15- and 30-minute time intervals. It is suitable for energy forecasting, renewable… See the full description on the dataset page: https://huggingface.co/datasets/mt0rm0/opsdata.OPSD-PI-SWE-Gym-512
OPSD-PI SWE-Gym Stage PI 512
Qwen3.5-9B stage-adaptive OPSD-PI 的公开 512-row 数据与 Weak from-scratch
训练包。Public 512-row data and Weak from-scratch training bundle.
Files
data/train.jsonl: 512 deterministic SWE-Gym rows with Weak, Medium, and
Strong PI for EXPLORE, REPRODUCE, DIAGNOSE, EDIT, and VERIFY.
data/manifest.json: source selection and integrity metadata.
release/OPSD_pi-opsd-pi-weak-from-scratch-20260818.tar.gz: immutable
source release containing launchers… See the full description on the dataset page: https://huggingface.co/datasets/LSW142857/OPSD-PI-SWE-Gym-512.opsd-probe-seed
OPSD prefix-continuation probe — seed data
Everything needed to reproduce the prefix-continuation probe for OPSD (on-policy
self-distillation) on a fresh GPU box, except the base model (Qwen/Qwen3-1.7B, pulled from
the Hub at setup) and the code repo (hbin0701/OPSD).
These artifacts live outside git because the training/eval output directory is .gitignored.
What the probe answers
Fitting p' = p + λ·(1[mode correct] − p) + γ against a properly sampled 64-shot… See the full description on the dataset page: https://huggingface.co/datasets/hbin0701/opsd-probe-seed.ops-dataopsd-plain-4b-rollouts
opsd-plain-4b-rollouts
This dataset contains rollout generations collected during training.
Source experiment
method: opsd-plain
model_size: 4b
experiment_dir: /home/irteam/outputs/opsd_plain_4b
Format
Each row contains:
step
sample_index
prompt
completion
method
model_size
source_file
Viewer structure
all: all rollout rows together
step_<N>: only one rollout step, easier to inspect in the dataset viewer
Notes… See the full description on the dataset page: https://huggingface.co/datasets/SeongryongJung/opsd-plain-4b-rollouts.OPSD-Example-Datadedup_filtered_HS3
Dedup-Filtered HelpSteer3 (+ rubrics on train)
A cleaned version of nvidia/HelpSteer3
(preference config) with three modifications:
Filter out domain == "multilingual" and overall_preference == 0 (tie) rows.
Within-split dedup by content hash (sha1(context, response1, response2)),
keeping the first occurrence. Upstream HS3 has ~35% byte-identical duplicate
rows after the filter step.
Cross-split dedup: drop validation rows whose content hash also appears
in train. Upstream HS3… See the full description on the dataset page: https://huggingface.co/datasets/opsd-genrm/dedup_filtered_HS3.opsd-plain-8b-rollouts
opsd-plain-8b-rollouts
This dataset contains rollout generations collected during training.
Source experiment
method: opsd-plain
model_size: 8b
experiment_dir: /home/irteam/outputs/opsd_plain_8b
Format
Each row contains:
step
sample_index
prompt
completion
method
model_size
source_file
Viewer structure
all: all rollout rows together
step_<N>: only one rollout step, easier to inspect in the dataset viewer
Notes… See the full description on the dataset page: https://huggingface.co/datasets/SeongryongJung/opsd-plain-8b-rollouts.olympiad_physics_stage1_qwen8b_opsdopsd_edge_dataset
OPSD Edge Dataset (Paper-Faithful)
Edge prompts for On-Policy Self-Distillation training on Nemotron-3-Nano.
Paper Reference
"Self-Distilled Reasoner: On-Policy Self-Distillation for LLMs"
Paper: arXiv:2601.18734
Code: github.com/siyan-zhao/OPSD
Dataset Description
This dataset contains 1,782 "edge" prompts where the 0.83 Nemotron adapter achieves 25-75% pass rate (uncertain cases ideal for learning).
Paper-Faithful Format
From Figure 2 of the OPSD… See the full description on the dataset page: https://huggingface.co/datasets/dvyomkesh/opsd_edge_dataset.aime25_stage1_qwen8b_opsdopsdc-training-dataOpenthoughts_math_30k_opsdaime26_stage1_qwen8b_opsdlcb_v5_stage1_qwen8b_opsdgpqa_diamond_stage1_qwen8b_opsdimo-answerbench_stage1_qwen8b_opsdhmmt-nov-2025_stage1_qwen8b_opsd