datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
SPSD-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.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-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-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.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.