datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
bactrainus-hotpotqa-teacher-traces
Bactrainus HotpotQA Teacher Traces
SOURCE-LINKED v1.0.0
Archived Llama 3.1 rationale and question-decomposition supervision, paired with complete SFT conversations and stable HotpotQA identities.
198,660 ROWS
4 CONFIGURATIONS
SFT MESSAGES
8B + 70B LABELS
CC BY-SA 4.0
A focused release of recovered teacher-generated supervision for multi-hop question answering. Every row contains the normalized annotation, an ordered… See the full description on the dataset page: https://huggingface.co/datasets/bactrianus/bactrainus-hotpotqa-teacher-traces.DeepSeek-V4-Flash-0731-Teacher-Distillation-40513x
DeepSeek V4 Flash 0731 Teacher Distillation — 40,513 Retained Rows
Teacher-distillation corpus generated with
deepseek-ai/DeepSeek-V4-Flash-0731.
The original manifest contained 45,000 unique seeds.
Following generation, QC, retry-based repair, quarantine auditing,
and recovery adjudication, 40,513 rows were retained.
Composition
Bucket
Rows
Coding
5,601
Agentic
9,982
Cyber blue
13,000
Controlled cyber red
6,999
Tool use
4,931
Total
40,513… See the full description on the dataset page: https://huggingface.co/datasets/trjxter/DeepSeek-V4-Flash-0731-Teacher-Distillation-40513x.qwen3-4b-teacher-rollouts-76k-nonthinking
Qwen3-4B Teacher Rollouts 76K Non-Thinking
This dataset contains 76,800 fixed teacher trajectories generated for a
prompt-aligned reproduction study of on-policy distillation with Qwen3-1.7B.
It is an independent research artifact, not an official release from the model
or paper authors.
Models and generation
Teacher: Qwen/Qwen3-4B-Instruct-2507
Tokenizer/chat template: Qwen/Qwen3-1.7B
Mode: non-thinking (enable_thinking=False)
Temperature: 0.7
Top-p: 1.0
Top-k:… See the full description on the dataset page: https://huggingface.co/datasets/YangyiH/qwen3-4b-teacher-rollouts-76k-nonthinking.stage3-real-expansion-agent-teacher-separated-pilot
Teacher-Separated Expansion Agent Pilot
A 10-task inspection batch generated by Qwen3-235B-A22B-Instruct-2507 from real
CLAPNQ, PubMedQA, MAUD, ContractNLI, and FinQA source tasks.
The teacher-only trajectory-generation system prompt is recorded in
metadata/generation-manifest.json for auditability, but is absent from every
saved training trajectory. Each final messages list begins with the real
memory-wrapped task user message, followed by native assistant expand calls,
exact… See the full description on the dataset page: https://huggingface.co/datasets/leonli66/stage3-real-expansion-agent-teacher-separated-pilot.lilm1-tool-teacher-corpora
LiLM1 tool teacher corpora
This dataset contains synthetic tool-use records generated with Gemma and Qwen
teacher models.
Method
Each teacher received structured tool schemas and task templates. One
configuration preserves the records from each teacher and task set.
Configurations
Configuration
Content
gemma-26b-a4b-function
Gemma function-calling records
qwen-27b-function
Qwen function-calling records
qwen-35b-a3b-function
Qwen MoE… See the full description on the dataset page: https://huggingface.co/datasets/glouriousgautam/lilm1-tool-teacher-corpora.deepscaler-teacher-sft-vllm-official-40k-clean-v2
DeepScaleR Teacher40k Clean v2
Filtered version of ThunderstormXXL/deepscaler-teacher-sft-vllm-official-40k.
Filtering
minimum official reward: 1.0
maximum text tokens: 8192
maximum response chars: 65000
near-duplicate SimHash hamming threshold: 4
required <think>...</think> and final boxed answer after reasoning
exact text/problem/response dedupe and near problem dedupe
Counts
raw examples: 40300
kept examples: 21727
train examples: 21292
val… See the full description on the dataset page: https://huggingface.co/datasets/ThunderstormXXL/deepscaler-teacher-sft-vllm-official-40k-clean-v2.gsm8k-qwen3.5-teacher-traces
GSM8K Qwen3.5 Teacher Traces
This dataset contains teacher-model reasoning traces and final answers generated with DashScope qwen3.5-397b-a17b for the official GSM8K train split from openai/gsm8k.
It was created as a reusable public artifact for research on mathematical reasoning, text-level distillation, filtering, and teacher-data analysis. The original GSM8K questions come from openai/gsm8k; this dataset adds generated teacher outputs and filtering metadata.
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/jerryjsjsj/gsm8k-qwen3.5-teacher-traces.rlve_teacher_topk16_20K
RLVE Teacher Top-16 Logit Data (20K)
Teacher top-k logit sidecar data for continuation-style KD-SFT warmup
(see compute_teacher_topk_logprobs.py / KDContinuationDataset).
Each row holds, per response token, the teacher's top-16 (+ forced true token)
candidate token ids and their log-probabilities, joined to the base dataset by
row_id.
Configs
rlve_offline_20K — 20,000 rows (rlve_offline_20K_teacher_top16.parquet)
rlve_rose_20K — 20,000 rows… See the full description on the dataset page: https://huggingface.co/datasets/CL-From-Nothing/rlve_teacher_topk16_20K.hanoi-crossing-aif-teacher
Hanoi Crossing AIF Teacher
25,000 unique prompt/completion examples generated by the observable GTSAM
active-inference controller on experiment/gtsam-aif-v2.
Each prompt contains only the acting player's visible poles and hand, legal
actions, remembered lift origin, and inferred opponent-state belief. It never
contains the referee's real opponent-private poles, hand, or private action.
The completion is one engine-legal action encoded as strict JSON.
Games cover N=1–4, both… See the full description on the dataset page: https://huggingface.co/datasets/anton96vice/hanoi-crossing-aif-teacher.deepscaler-teacher-sft-vllm-official-40k-clean-v3-no-reward-filter
deepscaler-teacher-sft-vllm-official-40k-clean-v3-no-reward-filter
Filtered version of ThunderstormXXL/deepscaler-teacher-sft-vllm-official-40k.
Filtering
reward filter enabled: False
minimum official reward: 1.0
scoring errors rejected: False
maximum text tokens: 8192
maximum response chars: 65000
near-duplicate SimHash hamming threshold: 4
required <think>...</think> and final boxed answer after reasoning
exact text/problem/response dedupe and near problem… See the full description on the dataset page: https://huggingface.co/datasets/ThunderstormXXL/deepscaler-teacher-sft-vllm-official-40k-clean-v3-no-reward-filter.deepscaler-teacher-sft-vllm-official-40k-clean-v4-conceptual
deepscaler-teacher-sft-vllm-official-40k-clean-v4-conceptual
Filtered version of ThunderstormXXL/deepscaler-teacher-sft-vllm-official-40k.
Filtering
reward filter enabled: False
minimum official reward: 1.0
scoring errors rejected: False
maximum text tokens: 32768
maximum response chars: 200000
near-duplicate SimHash hamming threshold: 4
required <think>...</think> and final boxed answer after reasoning
exact text/problem/response dedupe and near problem dedupe… See the full description on the dataset page: https://huggingface.co/datasets/ThunderstormXXL/deepscaler-teacher-sft-vllm-official-40k-clean-v4-conceptual.vgrout-leetcode-teacher-demos
vGROUT LeetCode teacher demonstrations
Cached teacher demonstrations used to warm up the
vGROUT gradient-routing experiments on the
ariahw/rl-rewardhacking LeetCode
environment. Each row is a full problem-specific completion. The kind column gives the
two demonstration types:
hack (215 rows): verified exploits of the run_tests loophole (hacked=True,
gt_pass=False).
solve (126 rows): correct solutions verified against the ground-truth tests
(gt_pass=True).
Why fewer… See the full description on the dataset page: https://huggingface.co/datasets/wassname/vgrout-leetcode-teacher-demos.