datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation.Qwen3.8-27B-Distillation-40K
Qwen3.8-27B-Distillation (40K Traces)
Qwen3.8-27B-Distillation is a dataset containing 40,000 reasoning traces distilled from Qwen's latest model — Qwen3.8-27B. We generated this dataset locally by running the model on our own infrastructure. It covers 4 domains with prompts sourced from 12 diverse open-source datasets.
Dataset Overview
Metric
Value
Total Examples
40,000
Teacher Model
Qwen3.8-27B
Model Precision
FP8
Reasoning Effort
medium… See the full description on the dataset page: https://huggingface.co/datasets/faunix/Qwen3.8-27B-Distillation-40K.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/o0Biggz0o/qwen3.8-max-glm5.2-kimi-k3-distillation.qwen3.8-max-distillation-50k
Qwen3.8-Max Distillation 50K
A curated dataset of 49,772 teacher-generated traces from qwen3.8-max-preview, prepared for supervised fine-tuning and off-policy knowledge distillation.
The teacher responses are preserved as returned by the API. Where the model emitted visible <think>...</think> blocks, those blocks remain in the assistant message. Some simpler prompts received direct answers without a thinking block.
[!CAUTION]
Terms and provenance notice — not cleared for… See the full description on the dataset page: https://huggingface.co/datasets/r0b0tlab/qwen3.8-max-distillation-50k.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/p-research/qwen3.8-max-glm5.2-kimi-k3-distillation.qwen3.8-max-glm5.2-kimi-k3-distill
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/ansulev/qwen3.8-max-glm5.2-kimi-k3-distill.qwen3.7-max-pi-tracesThis dataset was generated using teich by TeichAI
Prepare these datasets for supervised fine-tuning in just a few lines of code — see the Conversion section below.
Qwen3.7 Max Pi Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by qwen/qwen3.7-max.
JSONL files: 47
Training-ready tools
A complete configured tools schema snapshot is embedded in the collapsed section at the bottom of this README.
Use it… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/qwen3.7-max-pi-traces.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/inferenceport-ai/qwen3.8-max-glm5.2-kimi-k3-distillation.32B-reason-only.stride-1.k-8.statml-arxiv.qwen3-ids
32B-reason-only.stride-1.k-8.statml-arxiv.qwen3-ids
Thoughts for next-token prediction on k=8 token chunks of JackHsieh/statML-arxiv-40M-20M, generated by Qwen3-32B
with thinking mode off. Each thought is a few dense sentences of reasoning about the next
8 tokens after a cut, written from the document prefix alone — the generator never sees the
continuation. Stored thought_text includes the <thought>/</thought> wrapper.
The 4B parity counterpart is… See the full description on the dataset page: https://huggingface.co/datasets/JackHsieh/32B-reason-only.stride-1.k-8.statml-arxiv.qwen3-ids.maxrl_qwen3_4B_base_polaris_rollouts
MaxRL Qwen3-4B-Base training rollouts (POLARIS math prompts)
Every training rollout from an online RL run, with exact token ids, sampling
log-probs, and raw rewards — usable as a replay buffer to study off-policy RL
for LLM reasoning completely offline.
The run: Qwen3-4B-Base trained with the maxRL advantage estimator
(A = (r - mean)/(mean + eps), group mean over 16 rollouts per prompt;
maxRL paper) and a pure REINFORCE loss
(L = -A * log pi; no importance ratio, no clipping, no… See the full description on the dataset page: https://huggingface.co/datasets/ftajwar/maxrl_qwen3_4B_base_polaris_rollouts.4B-Instruct-reason-only.stride-1.k-8.statml-arxiv.qwen3-ids
4B-Instruct-reason-only.stride-1.k-8.statml-arxiv.qwen3-ids
Thoughts for next-token prediction on k=8 token chunks of JackHsieh/statML-arxiv-40M-20M, generated by
Qwen3-4B-Instruct-2507. Each thought is a few dense sentences of reasoning about the next
8 tokens after a cut, written from the document prefix alone — the generator never sees the
continuation. Stored thought_text includes the <thought>/</thought> wrapper.
This is the small-generator parity counterpart of… See the full description on the dataset page: https://huggingface.co/datasets/JackHsieh/4B-Instruct-reason-only.stride-1.k-8.statml-arxiv.qwen3-ids.qwen3.8-max-glm5.2-distillation-51389
Qwen3.8-Max / GLM-5.2 Distillation — 51,389 Rows
A deterministic, public Parquet release of admitted teacher traces for supervised fine-tuning, reasoning-format studies, tool-use studies, and tokenizer-specific rendering experiments. The sft configuration is the default training view. The package contains data and documentation only; it does not require executable dataset code.
Credits and Attribution
Dataset assembly and release packaging: r0b0tlab.
Qwen-derived… See the full description on the dataset page: https://huggingface.co/datasets/ufrik/qwen3.8-max-glm5.2-distillation-51389.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/bhadra123/qwen3.8-max-glm5.2-kimi-k3-distillation.openthoughts4-code-9168-prompts-qwen3-30b-a3b-thinking-2507-n16-flattened-logprobs-k16
OpenThoughts-4 Code SDG: Qwen3-30B-A3B-Thinking-2507 (n=16, top-16 logprobs)
Synthetic generations from
Qwen/Qwen3-30B-A3B-Thinking-2507
on the Marin OpenThoughts-4 code SDG prompt
set.
Each prompt is sampled n=16 times, and for every generated token the dataset
stores the chosen-token log probability plus the top-16 log probabilities
over the vocabulary, enabling distillation, KL-style fine-tuning,
reranking, and uncertainty analysis.
Generation setup
Field… See the full description on the dataset page: https://huggingface.co/datasets/marin-community/openthoughts4-code-9168-prompts-qwen3-30b-a3b-thinking-2507-n16-flattened-logprobs-k16.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/alliabba26/qwen3.8-max-glm5.2-kimi-k3-distillation.openthoughts4-code-9168-prompts-qwen3-32b-n16-flattened-logprobs-k16
OpenThoughts-4 Code SDG: Qwen3-32B (n=16, top-16 logprobs)
Synthetic generations from
Qwen/Qwen3-32B
on the Marin OpenThoughts-4 code SDG prompt
set.
Each prompt is sampled n=16 times, and for every generated token the dataset
stores the chosen-token log probability plus the top-16 log probabilities
over the vocabulary, enabling distillation, KL-style fine-tuning,
reranking, and uncertainty analysis.
Generation setup
Field
Value
Generator model… See the full description on the dataset page: https://huggingface.co/datasets/marin-community/openthoughts4-code-9168-prompts-qwen3-32b-n16-flattened-logprobs-k16.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/Distillio/qwen3.8-max-glm5.2-kimi-k3-distillation.qwen35-4b-drpo-vs0f49th-trainer-logprobs
Qwen3.5 4B DRPO trainer logprobs from W&B run vs0f49th
This dataset contains the raw trainer-logprob JSONL shards saved by W&B run ai2-llm/open_instruct_internal/vs0f49th (qwen35_4b_drpo__42__1782345587).
Contents
Source run: https://wandb.ai/ai2-llm/open_instruct_internal/runs/vs0f49th
Source path: /weka/oe-adapt-default/allennlp/deletable_rollouts/
Filename pattern: qwen35_4b_drpo__42__1782345587_trainer_logprobs_step*_rank*.jsonl
Files: 4320 JSONL shards… See the full description on the dataset page: https://huggingface.co/datasets/hamishivi/qwen35-4b-drpo-vs0f49th-trainer-logprobs.harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-30m-historical-20t-think
harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-30m-historical-20t-think
1,000 historical evaluation attempts (250 tasks, four samples per task), newly
graded with gpt-5.6-sol using Harvey's original per-criterion rubric prompt
and all-criteria-pass rule. Mean all-pass rate: 5.0000%.
The train split contains evaluation records, not training examples.
Generation and grading protocols
Generation is unchanged: historical 20-turn thinking-enabled
glob/grep/read agent… See the full description on the dataset page: https://huggingface.co/datasets/violetxi/harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-30m-historical-20t-think.openthoughts4-science-26041-prompts-qwen3-30b-a3B-thinking-2507-n8-flattened-logprobs-k16
OpenThoughts-4 Science SDG: Qwen3-30B-A3B-Thinking-2507 (n=8, top-16 logprobs)
Synthetic generations from
Qwen/Qwen3-30B-A3B-Thinking-2507
on the Marin OpenThoughts-4 science SDG prompt
set.
Each prompt is sampled n=8 times, and for every generated token the dataset
stores the chosen-token log probability plus the top-16 log probabilities
over the vocabulary, enabling distillation, KL-style fine-tuning,
reranking, and uncertainty analysis.
Generation setup
Field… See the full description on the dataset page: https://huggingface.co/datasets/marin-community/openthoughts4-science-26041-prompts-qwen3-30b-a3B-thinking-2507-n8-flattened-logprobs-k16.harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-3m-historical-20t-think
harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-3m-historical-20t-think
1,000 historical evaluation attempts (250 tasks, four samples per task), newly
graded with gpt-5.6-sol using Harvey's original per-criterion rubric prompt
and all-criteria-pass rule. Mean all-pass rate: 1.3000%.
The train split contains evaluation records, not training examples.
Generation and grading protocols
Generation is unchanged: historical 20-turn thinking-enabled
glob/grep/read agent… See the full description on the dataset page: https://huggingface.co/datasets/violetxi/harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-3m-historical-20t-think.harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-10m-historical-20t-think
harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-10m-historical-20t-think
1,000 historical evaluation attempts (250 tasks, four samples per task), newly
graded with gpt-5.6-sol using Harvey's original per-criterion rubric prompt
and all-criteria-pass rule. Mean all-pass rate: 4.0000%.
The train split contains evaluation records, not training examples.
Generation and grading protocols
Generation is unchanged: historical 20-turn thinking-enabled
glob/grep/read agent… See the full description on the dataset page: https://huggingface.co/datasets/violetxi/harvey-eval-gpt56sol-qwen35-9b-notes70-recall30-10m-historical-20t-think.financial-english-source-corpus-qwen35-1280
Financial English Source Corpus Qwen35 1280
This dataset is a filtered, fuzzy-deduplicated English source-text corpus for
financial-domain language-model training and translation-data generation. The
uploaded Parquet files are already prepared with the 1280-token source split
used by the downstream training pipeline.
This split version is derived from the pre-split
Financial English Source Corpus
by applying sentence-boundary splitting with the qwen3.5 tokenizer.… See the full description on the dataset page: https://huggingface.co/datasets/alwaysgood/financial-english-source-corpus-qwen35-1280.qwen3.8-max-glm5.2-kimi-k3-distillation
Multi-Teacher Distillation Dataset (57,937 traces)
A quality-filtered, deduplicated, multi-teacher SFT corpus combining traces from three frontier models across math, code, reasoning, instruction-following, tool-use, science, long-context, multilingual, and creative dialogue domains.
Teachers
Teacher
Provider
Traces
Qwen3.8-Max-Preview
Alibaba Cloud Model Studio
48,283
GLM-5.2
Z.AI Coding Plan
5,307
Kimi Code K3
Moonshot AI (Kimi)
4,347… See the full description on the dataset page: https://huggingface.co/datasets/Lalo42/qwen3.8-max-glm5.2-kimi-k3-distillation.qwen3-5-tiny-cpu-repro-v1
Qwen3.5 tiny native random CPU fixture
Complete randomly initialized, untrained Qwen3_5ForConditionalGeneration checkpoint.
This is a pipeline/reproducibility fixture, not a useful language model, distillation,
quantization, quality benchmark, or claim about the performance of Qwen3.8-27B.
No upstream model weights or training data were used. No paid GPU/cloud compute.
Architecture and lineage
Architecture lineage: Qwen/Qwen3.8-27B at… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qwen3-5-tiny-cpu-repro-v1.qwen3-8b-aime-2009-2024-16x
AIME Reasoning Traces · Qwen3-8B
7,680 reasoning traces for 480 AIME problems, with 16 sampled responses per problem. The corpus covers AIME I and AIME II from 2009 through 2024 and includes both correct and incorrect answers.
We created this dataset for How Should Incorrect Traces Be Used in Supervised Fine-Tuning? It is the source for the paper's larger AIME experiment, which compares correct and incorrect supervision across two disjoint sets of 106 problems.
Code and… See the full description on the dataset page: https://huggingface.co/datasets/suryadv/qwen3-8b-aime-2009-2024-16x.Qwen3-Coder-Next-OpenCode-Preference
Dataset Card — OpenCode Rejection Sampling (Preference)
Overview
This dataset contains 10,920 preference pairs for preference-based training (DPO, KTO, SimPO, ORPO, etc.) on competitive programming tasks. Each pair consists of:
Chosen: a candidate solution that passes 100% of test cases
Rejected: a candidate solution that fails, with a fine-grained rejection type label
Pairs are produced via rejection sampling with Qwen3-Coder-Next: 8 candidate solutions are… See the full description on the dataset page: https://huggingface.co/datasets/zake7749/Qwen3-Coder-Next-OpenCode-Preference.openthoughts4-science-26041-prompts-qwen3-32b-n8-flattened-logprobs-k16
OpenThoughts-4 Science SDG: Qwen3-32B (n=8, top-16 logprobs)
Synthetic generations from
Qwen/Qwen3-32B
on the Marin OpenThoughts-4 science SDG prompt
set.
Each prompt is sampled n=8 times, and for every generated token the dataset
stores the chosen-token log probability plus the top-16 log probabilities
over the vocabulary, enabling distillation, KL-style fine-tuning,
reranking, and uncertainty analysis.
Generation setup
Field
Value
Generator model… See the full description on the dataset page: https://huggingface.co/datasets/marin-community/openthoughts4-science-26041-prompts-qwen3-32b-n8-flattened-logprobs-k16.Qwen3-Coder-Next-Open-Code-SFT
Dataset Card — OpenCode Rejection Sampling
Overview
This dataset contains high-quality code reasoning data for training language models on competitive programming tasks. It is produced via rejection sampling with Qwen3-Coder-Next, which would generate multiple candidate solutions per problem, each candidate is executed against test cases in a sandboxed environment, and the results are used to build two complementary training datasets:
SFT dataset (49,374 examples)… See the full description on the dataset page: https://huggingface.co/datasets/zake7749/Qwen3-Coder-Next-Open-Code-SFT.Qwen3-Math-Eval
Qwen3 Math Evaluation Suite
Greedy (temperature 0) outputs from Qwen3 1.7B / 4B / 8B / 14B on nine math-reasoning benchmarks across output-token budgets {2k, 4k, 8k, 16k, 32k}. 1,417,388 predictions over 186 model-by-dataset-by-budget cells, each with the full reasoning trace, the extracted answer, and strict and answer-forced correctness labels.
On standard MATH (MATH-500, Hendrycks MATH test, competition MATH) these models are saturated at 16k: the 4B is at or above 0.94 and… See the full description on the dataset page: https://huggingface.co/datasets/ssubhnil/Qwen3-Math-Eval.