datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
proof-pile-2-fixed
The original EleutherAI/proof-pile-2 dataset uses a custom python script and .jsonl.zst files, which some versions of the datasets library struggle with.
This dataset contains the same data, subsets, and splits as EleutherAI/proof-pile-2, converted into standard parquet format.
Each subset and split was also shuffled so that you can directly train on the data without issue.
Conversion was performed using the following script:
import os
importzstandard as zstd
import json
import pandas as pd… See the full description on the dataset page: https://huggingface.co/datasets/aklein4/proof-pile-2-fixed.hermes3-en-fixed
Dataset Card for Hermes 3 Fixed Conversations
Dataset Description
Dataset Summary
hermes3-en-fixed is a [NousResearch/Hermes-3-Dataset]. During preparation we removed all system prompts and normalized the message roles and content to match the common schema we use across our dialog datasets.
Languages
English (en)
Dataset Structure
Data Fields
conversations: list of messages in a dialog (array of objects)
from: normalized sender role — user or assistant… See the full description on the dataset page: https://huggingface.co/datasets/lapa-llm/hermes3-en-fixed.TheBlueScrubs-v1-fixed
openmed-community/TheBlueScrubs-v1-fixed
What is this?
TheBlueScrubs-v1-fixed is a maintenance fork of the upstream TheBlueScrubs/TheBlueScrubs-v1 train split that resolves a schema bug in the meta column.In the original train files, some rows serialized meta incorrectly (appearing as the literal string "dict"). This fork re-exports the entire train split without meta column, preserving text field and values.
Document count: 11,080,331 texts (train)
Tokens (upstream… See the full description on the dataset page: https://huggingface.co/datasets/openmed-community/TheBlueScrubs-v1-fixed.2026-07-29-msm-philosophy-spec-fixed-eval
Fixed evaluation: does model-spec midtraining change harmful-omission or provenance behaviour?
experiment: Byte-identical single-turn fixed evaluation across seven matched checkpoints, designed to attribute (or rule out) an effect of model-spec midtraining (MSM) on two behaviours: treating tool-channel content as an instruction (prov-* probes) and suppressing a warranted safety concern under instruction (omis-* probes). This is the attribution step behind the investigation's… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-07-29-msm-philosophy-spec-fixed-eval.fixed-kkc-dataset
Fixed KKC Dataset
日本語Wikipedia入力誤りデータセット (v2) から生成した、かな漢字変換（KKC）タスク用の選好ペアデータセットです。
データセットの概要
Wikipediaの編集差分のうち kanji-conversion_a カテゴリ（誤変換の修正）に該当するものを抽出しています。
各レコードは、カタカナの読みに対して「正しい漢字表記（chosen）」と「誤った表記（rejected）」のペアを持ちます。
かな漢字変換モデルの学習・評価や、選好学習（RLHF / DPO）に利用できます。
データ形式
各レコードは以下のフィールドを持つ JSON Lines 形式です。
フィールド
型
説明
left_context
string
変換箇所より前の文脈テキスト
prompt
string
変換対象語のカタカナ読み
chosen
string
正しい漢字表記（Wikipedia編集後）
rejected
string… See the full description on the dataset page: https://huggingface.co/datasets/yuuki14202028/fixed-kkc-dataset.Qwen3.7_5k_fr60_fixed
Qwen 3.7 Max Thinking — Distilled Reasoning Dataset (FR60, cleaned)
5,000 chain-of-thought (CoT) reasoning traces, ~60% machine-translated to French, derived from the original dataset WithinUsAI/Qwen3.7_Max_Thinking_dataset_5K.
Each example contains a problem, a detailed step-by-step reasoning trace (in the Qwen 3.7 Max Thinking style), and a concise final answer.
Source and translation
This dataset is a partial translation of the original English dataset… See the full description on the dataset page: https://huggingface.co/datasets/Tivaphraen/Qwen3.7_5k_fr60_fixed.TRQA-fixed
TRQA (fixed configs)
Private convenience mirror of GENTEL-Lab/TRQA, with each
schema exposed as a separate Hugging Face dataset configuration so that Dataset Viewer and load_dataset work.
The CSV contents are unchanged from source revision
c712c7948c907dec61beada11951cf997d89bae4.
Config
Rows
Columns
Default
lit-choice
172
Question, Options, Answer
Yes
lit-short
1,108
Question, Answer
No
db
641
Question, Answer
No
from datasets import load_dataset
choice =… See the full description on the dataset page: https://huggingface.co/datasets/haydn-jones/TRQA-fixed.Dolci-Instruct-SFT-Tool-Use-Fixed
Dolci-Instruct-SFT-Tool-Use-Fixed
Dataset Description
Dolci-Instruct-SFT-Tool-Use-Fixed is a cleaned and re-formatted version of the allenai/Dolci-Instruct-SFT-Tool-Use tool-use dataset. It is designed as the tool-calling (function-calling) extension of the openbmb/UltraData-SFT-2605 Supervised Fine-Tuning dataset, so that tool-use samples can be mixed into UltraData-SFT-2605 training runs seamlessly.
The raw Dolci-Instruct-SFT-Tool-Use data uses a custom message… See the full description on the dataset page: https://huggingface.co/datasets/nekocyrene/Dolci-Instruct-SFT-Tool-Use-Fixed.orca-agentinstruct-1M-v1-cleaned-fixed-sharegpt
Orca Agentic Instruct - ShareGPT
Entries with empty assistant turn were removed
messages were converted into ShareGPT
Initial number of examples: 1046410
Number of examples removed: 68
Remaining examples: 1046342
Removal rate: 0.01%
Credit
mlabonne/orca-agentinstruct-1M-v1-cleaned
blended-skill-talk-fixed
Compatibility Update
This repository is a compatibility-fixed version of the original Blended Skill Talk dataset.
The original dataset can be found at:
Original Hugging Face dataset: https://huggingface.co/datasets/anezatra/blended-skill-talk
This version was created to maintain compatibility with newer versions of the Hugging Face datasets library.
Changes from the Original Dataset
The following changes were made:
Removed the unused label_candidates column.… See the full description on the dataset page: https://huggingface.co/datasets/TutorialGuide/blended-skill-talk-fixed.icd-11-qa-fixedA fixed version from the original Lamini ICD-11 QA Dataset
Bluemoon_Top50MB_Sorted_Fixed_ja
Bluemoon_Top50MB_Sorted_Fixed_ja
SicariusSicariiStuff/Bluemoon_Top50MB_Sorted_Fixedを、GENIAC-Team-Ozaki/karakuri-lm-8x7b-chat-v0.1-awqを用いて日本語に翻訳したロールプレイ学習用データセットです。
LLMの推論にはDeepInfraというサービスを使いました。
翻訳の詳細
3-shots promptingでの翻訳
mistralのtokenizerで出力が8000トークンを超えるまで翻訳
元データセットにある非常に長い対話は上記条件で途中のターンで翻訳を終了しています。
LLM特有の同じ出力が繰り返される現象に遭遇した場合、その時点で該当レコードの翻訳を終了
この結果1ターン未満となったレコード（157件）を削除… See the full description on the dataset page: https://huggingface.co/datasets/Aratako/Bluemoon_Top50MB_Sorted_Fixed_ja.qpl-multi-decomposer-fixed-ds
QPL Multi-Decomposer Dataset — Fixed
A manually reviewed and cleaned version of d4nieldev/qpl-multi-decomposer-ds, with invalid decompositions removed and their chain-of-thought reasoning blocks rewritten as rejection explanations.
Relationship to the original dataset
Train split — manually reviewed and fixed (see Changes from the original below).
Validation split — identical to the validation split of d4nieldev/qpl-multi-decomposer-ds, unchanged. It was not part of the… See the full description on the dataset page: https://huggingface.co/datasets/d4nieldev/qpl-multi-decomposer-fixed-ds.batch_test_fixed
Qwen Continuation Dataset
Generated with qwen_continuation_dataset.
Statistics
Shards
6
Examples
52
Shard size
10
Updated
2026-07-13 10:07 UTC
Usage
from datasets import load_dataset
ds = load_dataset("TeoStarshine/batch_test_fixed")
ds = load_dataset("TeoStarshine/batch_test_fixed", streaming=True)
Fields
Field
Description
source_id
source document ID
source_name
source dataset (fineweb / math)… See the full description on the dataset page: https://huggingface.co/datasets/TeoStarshine/batch_test_fixed.mo9_b_fixed_monitor_output
MO9 Atlas-9 v2 — M6 Monitor Format
Training dataset for the MO9 Sleeper Agents replication. This version updates the monitor user prompt to the M6 format, which asks for a 1-sentence explanation before the verdict.
Files
File
Records
Description
train_36k_combined.jsonl
36,000
Shuffled combination of policy (18k) + monitor (18k). Ready for training.
train_18k_monitor_m6.jsonl
18,000
Monitor component only (M6 format).
Shuffle seed: 42
What… See the full description on the dataset page: https://huggingface.co/datasets/jprivera44/mo9_b_fixed_monitor_output.Atcgpt-Fixed2
Dataset Card for Atcgpt-Fixed2
This dataset contains instruction-input-output pairs converted to ShareGPT format, designed for instruction tuning and text generation tasks.
Dataset Description
The dataset consists of carefully curated instruction-input-output pairs, formatted for conversational AI training. Each entry contains:
An instruction that specifies the task
An optional input providing context
A detailed output that addresses the instruction
Usage… See the full description on the dataset page: https://huggingface.co/datasets/HappyAIUser/Atcgpt-Fixed2.turkish-gsm8k-fixed-15k
⚠️ Deprecated / Kullanmayın
Bu dataset deprecated durumdadır ve yeni eğitimlerde kullanılmamalıdır.
Sebep
Deprecated: deep audit found step-level arithmetic/logic errors in ~28.3% of examples. Use the verified UTF-8 math replacement instead.
Yerine kullanın
Genel Türkçe SFT: https://huggingface.co/datasets/kilicai/turkish-sft-clean-v2-balanced
Doğrulanmış Türkçe matematik/CoT: https://huggingface.co/datasets/kilicai/turkish-gsm8k-perfect-tr-15k
Not… See the full description on the dataset page: https://huggingface.co/datasets/kilicai/turkish-gsm8k-fixed-15k.ssd-math-v1p1-e06m-fixed-batch-16
SSD Math V1.1-E06M fixed batch
Tiny 16-record SFT JSONL fixture used for the V1.1-E06M 4B single-batch LR diagnostic.
The records are generated math-reasoning traces from the local v0 overfit fixture artifacts/overfit/v0_unique_batch_16.jsonl.
cs8-fixed
cs8-fixed
CS8 fixed dataset — deterministic inner alias assignment, used for circuit analysis and SAE training
Structure
train.jsonl — 12,000 training examples (20 aliases × 600)
val.jsonl — 12,000 validation examples (20 aliases × 600)
alias_vocab.json — 20 aliases with T1/T2 tokenization group labels
Alias Groups
T1 (single-token): emp, inv, txn, mgr, ord, prod, cust, dept, acct, sale
T2 (two-token, generic first subtoken): shp, whs, rgn, lne… See the full description on the dataset page: https://huggingface.co/datasets/Likithp/cs8-fixed.ATCgpt-Fixed
Dataset Card for ATCgpt-Fixed
This dataset contains instruction-input-output pairs converted to ShareGPT format, designed for instruction tuning and text generation tasks.
Dataset Description
The dataset consists of carefully curated instruction-input-output pairs, formatted for conversational AI training. Each entry contains:
An instruction that specifies the task
An optional input providing context
A detailed output that addresses the instruction
Usage
This… See the full description on the dataset page: https://huggingface.co/datasets/HappyAIUser/ATCgpt-Fixed.nobatched_test_fixed
Qwen Continuation Dataset
Generated with qwen_continuation_dataset.
Statistics
Shards
5
Examples
50
Shard size
10
Updated
2026-07-13 10:26 UTC
Usage
from datasets import load_dataset
ds = load_dataset("TeoStarshine/nobatched_test_fixed")
ds = load_dataset("TeoStarshine/nobatched_test_fixed", 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/TeoStarshine/nobatched_test_fixed.batch_test_fixed8
Qwen Continuation Dataset
Generated with qwen_continuation_dataset.
Statistics
Shards
5
Examples
50
Shard size
10
Updated
2026-07-13 10:40 UTC
Usage
from datasets import load_dataset
ds = load_dataset("TeoStarshine/batch_test_fixed8")
ds = load_dataset("TeoStarshine/batch_test_fixed8", 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/TeoStarshine/batch_test_fixed8.