datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
qasr-arabic-speech-continuationsdsg-state-continuation
DSG State-Continuation
Training data for graph-conditioned long-form fiction generation: given the
narrative state a reader would hold after chapters 1..t-1, and a one-line brief
for chapter t, write chapter t.
Built from 215 public-domain novels (Project Gutenberg, English fiction),
segmented into chapters. 6,876 examples.
Why the state is built this way
The state is not a summary and not a retrieval index. It is a revision-aware
assertion store built causally —… See the full description on the dataset page: https://huggingface.co/datasets/GOVINDFROM/dsg-state-continuation.qwen_continuation_dataset
Qwen Continuation Dataset
Generated with qwen_continuation_dataset.
Statistics
Shards
42
Examples
235
Shard size
1
Updated
2026-07-05 14:47 UTC
Usage
from datasets import load_dataset
ds = load_dataset("Zhuzhik/qwen_continuation_dataset")
ds = load_dataset("Zhuzhik/qwen_continuation_dataset", streaming=True)
Fields
Field
Description
source_id
source document ID
source_name
source dataset (fineweb… See the full description on the dataset page: https://huggingface.co/datasets/Zhuzhik/qwen_continuation_dataset.CLAMP-Sampled-Continuations-and-Demos
CLAMP Sampled Continuations and VLABench Demos
This dataset accompanies HLR/CLAMP, the Constrained Language-Action Model Planner.
It contains 20 closed-loop VLABench case studies: 10 successful and 10 unsuccessful executions. Each case includes the input image and mask, instruction, prompt and entity metadata, CLAMP and ground-truth plans, evaluation output, execution video, and a frame manifest.
Layout
demo_dataset/
├── manifest.json
├── README.md
├──… See the full description on the dataset page: https://huggingface.co/datasets/SueMintony/CLAMP-Sampled-Continuations-and-Demos.dream-of-the-red-chamber-continuations
红楼梦续写 · Dream of the Red Chamber: 100 AI Continuations
项目简介
本数据集包含 92 个独立的AI续写版本，续写中国古典文学巅峰之作《红楼梦》的第八十一回至第一百零八回（共28回）。所有续写严格遵循曹雪芹前八十回中埋下的伏笔、谶语和人物命运，完全拒绝高鹗续书。
为什么做这个数据集
《红楼梦》的结局是世界文学史上最大的悬案之一。曹雪芹约于1763年去世前未能完成全书，仅留下前八十回。1791年左右，高鹗发表了一百二十回本，补写了后四十回，但红学研究日益表明高鹗续书严重违背了曹雪芹在前八十回中精心布置的伏笔。
曹雪芹原意 vs 高鹗续书
情节
曹雪芹原意
高鹗续书
黛玉之死
泪尽而亡，呼应"绛珠还泪"神话
焚稿断痴情
宝玉宝钗婚姻
"纵然是齐眉举案，到底意难平"
掉包计骗婚
贾府败落
政治牵连，锦衣军抄家，"忽喇喇似大厦倾"
败而复兴，"兰桂齐芳"
结局… See the full description on the dataset page: https://huggingface.co/datasets/PursuitOfDataScience/dream-of-the-red-chamber-continuations.all_continuations_sentence_doubt_distchinese-novel-continuation-precise-tokens
中文小说续写精确 Token 长度数据集
基于金庸《神雕侠侣》的高精度中文文本生成数据集，专为 GPU 内存测试和序列长度性能分析而设计。
数据集特点
高精度：99.4%+ 的目标 token 长度准确率
多种长度：5 个变体（1024、2048、4096、8192、16384 tokens）
统一格式：Alpaca 格式的小说续写任务
质量控制：95%+ 样本在目标长度的 ±2% 范围内
数据集统计
目标 Tokens
实际平均
准确率
样本数
文件大小
±1% 内样本
±2% 内样本
1024
1017.9
99.4%
800
2.5MB
739
775
2048
2037.7
99.5%
800
4.9MB
764
795
4096
4078.1
99.6%
800
9.8MB
792
800
8192
8158.1
99.6%
800
19.5MB
800
800
16384
16317.6
99.6%
800
39.1MB
800
800… See the full description on the dataset page: https://huggingface.co/datasets/aweffr/chinese-novel-continuation-precise-tokens.all_continuations_sentence_doubt_dist_doubtful_sentences_embedded_doubtful_sentencesreal_toxicity_continuationsall_wrong_continuations_sentence_doubt_distembeddings_all_correct_continuationsnyt_texts_filtered_prompt_continuationryan-greenblatt-sft-continuation
Ryan Greenblatt simulator — continuation format
Direct continuation. Prompt = post title + opening (or recent) paragraphs of a Ryan Greenblatt LessWrong post. Target = the next ≤1500 cl100k tokens of paragraph-aligned prose. Long posts are chunked into multiple overlapping rows so the full post is reachable.
Part of the Ryan-Greenblatt-simulator SFT project (Tinker / Qwen3-8B-base).
See OVERALL_PLAN.md and proposal.md in the source repo for context.
Splits
train: 251… See the full description on the dataset page: https://huggingface.co/datasets/abhayesian/ryan-greenblatt-sft-continuation.self-critiquing-base-selected-900-prompt-continuation
Dataset Card for "self-critiquing-base-selected-900-prompt-continuation"
More Information needed
clinical-meta-narrative_continuations_v0.4Meta Clinical Narrative Continuations v0.4
What this dataset tests
Whether a model interrupts narrative continuation when the evidence shifts.
This version adds
timeline steps
cross trial continuation traps
audience shift pressure
submission positioning
Input columns
data_anchor
continuation_draft
timeline_step
audience
claim_ceiling_allowed
submission_position_expected
Model task
Return one JSON object
breacheslist of breach labels
corrected_continuationone paragraph… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-meta-narrative_continuations_v0.4.Continuation-egy-for-ultravox-v1
Speech Dataset with Continuations
This dataset contains speech audio files with their transcriptions and AI-generated continuations.
Dataset Splits
Train: 85,248 samples (70.0%)
Validation: 18,267 samples (15.0%)
Test: 18,268 samples (15.0%)
Dataset Structure
Each sample contains:
audio: Audio file (WAV format, 16kHz)
text: Original transcription text
continuation: AI-generated continuation of the text
duration: Audio duration in seconds
sampling_rate_hz:… See the full description on the dataset page: https://huggingface.co/datasets/MAdel121/Continuation-egy-for-ultravox-v1.labels_all_continuations_sentencessingaporean_accent_district_names_continuationtable-continuation-datasetgigaspeech-seed-context-continuation-noiseself-critiquing-refine-continuations
Dataset Card for "self-critiquing-refine-continuations"
More Information needed
embeddings_all_wrong_continuationsdl_hw2_text_detector_data_continuation_patch_v1qwen35-continuation-bench
Qwen Continuation Dataset
Generated with qwen_continuation_dataset.
Statistics
Shards
1
Examples
100
Shard size
500
Updated
2026-07-09 18:29 UTC
Usage
from datasets import load_dataset
ds = load_dataset("TeoStarshine/qwen35-continuation-bench")
ds = load_dataset("TeoStarshine/qwen35-continuation-bench", streaming=True)
Fields
Field
Description
source_id
source document ID
source_name
source… See the full description on the dataset page: https://huggingface.co/datasets/TeoStarshine/qwen35-continuation-bench.ContinuationDatasetADI17-continuationself-critiquing-critique-continuations
Dataset Card for "self-critiquing-critique-continuations"
More Information needed
meta-clinical-narrative-continuations-v0.3
Clinical Narrative Continuations
Meta Dataset v0.3
Purpose
This dataset tests whether a model:
Stops when the clinical narrative stops
Resists completing truncated notes
Avoids inventing downstream events
Preserves narrative boundaries under incomplete input
You are testing restraint against narrative momentum.
Why this matters
Clinical documentation is often partial.
Notes are cut off.Dictations end mid-sentence.Handovers omit outcomes.… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/meta-clinical-narrative-continuations-v0.3.qwen_continuation_dataset2
Qwen Continuation Dataset
Generated with qwen_continuation_dataset.
Statistics
Shards
2
Examples
20
Shard size
10
Updated
2026-07-12 10:41 UTC
Usage
from datasets import load_dataset
ds = load_dataset("TeoStarshine/qwen_continuation_dataset2")
ds = load_dataset("TeoStarshine/qwen_continuation_dataset2", streaming=True)
Fields
Field
Description
source_id
source document ID
source_name
source… See the full description on the dataset page: https://huggingface.co/datasets/TeoStarshine/qwen_continuation_dataset2.all_wrong_continuations