datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
KIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/KIMI-K2.5-1000000x.GLM-5.2-Conversation
GLM-5.2 · Conversation-50000x
50,000x traces distilled from GLM-5.2 on High reasoning
Token Count: 120M
Distribution:
Speaking domains:
•Greetings
•Customer Support
•Step by step explanations
•Motivational language
•Logical Questions
•Creative Writing
STEM:
•Algebra, calculus, quantum mechanics concepts
•Astromony and astrophysics
•Datascience and machine learning
•Biology
Programming:… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/GLM-5.2-Conversation.GLM-5.2-Finance-80000x
GLM-5.2 · Finance-80000x
80,000x financial related traces distilled from GLM-5.2 on High reasoning
Risk · Markets · Investments · Corporate Finance · Wealth Management
Token Count: 220M
Unique prompts generated with diffusion Gemma-27B answered by GLM-5.2
You can use this dataset for any purpose and you dont need to credit me, preferably dont claim it as your own.
hi - ianncity
GLM-5.2-Logic-Puzzles
GLM-5.2 · Logical Puzzles
6000x traces distilled from GLM-5.2 on High reasoning
Token Count: 5M~?
Distribution:
Puzzles:
•Tokenization blindless ex: counting the r's in strawberry
•Goal reasoning ex: the car wash test (theres no car wash question exactly just prompts like it so its not just benchmaxxing)
•Reading comprehension traps
•Temporal reasoning
•Many other categories not worth mentioning
Prompts… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/GLM-5.2-Logic-Puzzles.Fable-5-traces
Glint Research Dataset Card
Fable 5 Pi Agent Traces
A compact, high-signal corpus of Fable 5 coding-agent traces converted into Hugging Face Agent Traces / Pi-compatible sessions for Data Studio inspection, tool-use policy learning, and reasoning/action distillation.
Primary Config
pi_agent/train
Agent Trace preview enabled
4,665 Pi trace sessions
60 source sessions
3,799 tool… See the full description on the dataset page: https://huggingface.co/datasets/IAMRonHIT/Fable-5-traces.GLM-5.2-Science
GLM-5.2 · Science-50000x
50,000x traces distilled from GLM-5.2 on High reasoning
Physics · Chemistry · Biology
Token Count: 160M
Theres prompt overlap with my Kimi K2.5 dataset science subset, which I think those prompts are getting used in alot of places now
You can use this dataset for any purpose and you dont need to credit me, preferably dont claim it as your own.
hi - ianncity
ui-instruct-4k
UI Instruct 4K
A instruction-completion dataset for finetuning language models to specialize in generating Next.js / ShadCN UI components using React, TypeScript, and Tailwind CSS.
Dataset Summary
This dataset was created with the primary goal of finetuning Qwen 3.5 4B to become a specialist at outputting production-ready Next.js and ShadCN-based UI components. Each example consists of a natural language prompt describing a UI component or layout, paired with a clean… See the full description on the dataset page: https://huggingface.co/datasets/iamdyeus/ui-instruct-4k.I-ARIF-CANON
17
🗺️ Position in I-ARIF Governance Stack
This dataset is part of the arifOS constitutional governance training-and-evaluation pipeline — a closed-loop alignment substrate.
#
Dataset
Role
Downloads
License
1
AAA
Constitutional substrate — doctrine + gold eval
161
AGPL-3.0
2
BBB
Baseline behavior benchmark — ILMU API audit
247
CC-BY-4.0
3
CCC
Alignment contrast corpus — ILMU vs kernel
193
CC-BY-4.0
4
DDD
Register-sensitivity probe — Penang loghat… See the full description on the dataset page: https://huggingface.co/datasets/ariffazil/I-ARIF-CANON.us-caselaw-ia
Iowa Case Law
Code & tools: github.com/docketx — legal-scrambler pseudonymises a case file on your own hardware before a frontier model sees it; claude-for-legal is the Claude Code plugin (docketx-open-law) that loads these datasets and checks citations against them.
Source & credit — Free Law Project / CourtListener
Every opinion in this dataset comes from the Free Law Project / CourtListener bulk export of
2026-06-30 (10,798,347 opinions). CourtListener… See the full description on the dataset page: https://huggingface.co/datasets/docketx/us-caselaw-ia.nepali_news_textpokerbench-8max-reasoning-traces
PokerBench 8-max — teacher-distilled reasoning traces
Reasoning traces for 8-max No-Limit Hold'em decisions, distilled from Claude
Sonnet 5 on Bedrock in the STaR style, for training small models to reason
about poker prices rather than pattern-match to an action.
Method
The teacher is not told the answer. It reasons freely from the same prompt
production sends, and a trace is kept only if its conclusion matches the target
label. Telling the teacher the target… See the full description on the dataset page: https://huggingface.co/datasets/ianlee1996/pokerbench-8max-reasoning-traces.pokerbench-rl-dpo
PokerBench RL — Counterfactual DPO Preference Data
DPO preference pairs and raw self-play logs for training a Texas Hold'em LLM
to exploit non-GTO opponents, addressing the PokerBench paper's Future Work observation that pure SFT models lose to GPT-4-style "donking" strategies.
This dataset feeds the ianlee1996/pokerbench-qwen3-14b-lora-dpo checkpoint training.
How it was built
Self-play (5000 hands): ianlee1996/pokerbench-qwen3-14b-lora-mixed (Qwen3-14B + LoRA… See the full description on the dataset page: https://huggingface.co/datasets/ianlee1996/pokerbench-rl-dpo.gpt-oss-20b-reasoning-traces
GPT-OSS-20B Reasoning Traces
3,333 reasoning traces generated by openai/gpt-oss-20b and filtered for clean, terminating reasoning. It was built to distill GPT-OSS's tight reasoning style into smaller models, and is the training set behind iAmBoosted/Qwen3.5-9B-OSS-Distilled.
What's in it
Each record pairs a prompt with GPT-OSS-20B's full reasoning trace and final answer, in chat-message form, ready for supervised fine-tuning (SFT).
~4,000 raw traces were generated, then… See the full description on the dataset page: https://huggingface.co/datasets/iAmBoosted/gpt-oss-20b-reasoning-traces.execution-verified-agent-trajectories
Execution-Verified Agent Trajectories — Format & Method
This repository documents a method and data format for building supervised fine-tuning sets from agent
trajectories that are verified by running the code, not by asking a model whether the answer looks right.
This is a specification plus synthetic examples, not a corpus. The trajectories that trained
Luthor 8B were generated against a private repository and cannot be
released. Everything needed to rebuild an equivalent set… See the full description on the dataset page: https://huggingface.co/datasets/IAMIbrahim/execution-verified-agent-trajectories.crisis-response-training-v2
Crisis Response Training Dataset
A synthetic dataset of 2,000 training examples for fine-tuning language models on crisis response scenarios. Each example includes structured responses from both civilian and first responder perspectives.
Dataset Description
This dataset contains 2,000 instruction examples in Unsloth Alpaca format, generated synthetically using large language models (LLMs) for training crisis response systems. The data is designed to help models learn… See the full description on the dataset page: https://huggingface.co/datasets/ianktoo/crisis-response-training-v2.IALA_B
IALA_B: Daytime, Single‑Mark JSON→JSON Dataset (1,500 samples)
Input = VLM observation JSON (body/topmark attributes; may include synonyms/unknown).Output = LLM normalized JSON (mark_type, safe_direction, hazards, consistency, finalized_attributes…).Lateral mapping assumes IALA‑B (red=starboard, green=port).
Total samples: 1500
Classes (9): cardinal_east, isolated_danger, lateral_port, lateral_starboard, cardinal_north, safe_water, cardinal_south, special_mark, cardinal_west… See the full description on the dataset page: https://huggingface.co/datasets/gabrielsunhyuck/IALA_B.MediFlowThinks
MediFlow
A large-scale synthetic instruction dataset of 2.5M rows (~700k unique instructions) for clinical natural language processing covering 14 task types and 98 fine-grained input clinical documents.
t-SNE 2D Plot of MediFlow Embeddings by Task Types
Dataset Splits
mediflow: 2.5M instruction data for SFT alignment.
mediflow_dpo: ~135k top-quality instructions with GPT-4o generated rejected_output for DPO alignment.
Main Columns
instruction:… See the full description on the dataset page: https://huggingface.co/datasets/IAMRonHIT/MediFlowThinks.gemma-3-12b-it-lmsys-onpolicy-rollouts
On-policy chat rollouts: google/gemma-3-12b-it on LMSYS-Chat-1M prompts
Each row is a first-user-turn prompt sampled from
lmsys/lmsys-chat-1m and a
response generated on-policy by google/gemma-3-12b-it with vLLM (do_sample,
temperature 0.7, top_p 1.0, max_new_tokens 768, seed 42). 24,991 rows. Built to match
GemmaScope 2's instruction-tuned SAE training distribution (real model rollouts) for a
short KL+MSE ("end-to-end") finetune of the released GemmaScope-2 residual SAE.… See the full description on the dataset page: https://huggingface.co/datasets/iarcuschin/gemma-3-12b-it-lmsys-onpolicy-rollouts.fable-code-distill
Fable Code Distillation Dataset
348 high-quality instruction/response pairs distilled from Fable for training a
purpose-built ternary code agent (student: PrismML Ternary-Bonsai).
Domains: Python (OOP, async/concurrency), JavaScript/TypeScript, bash/shell, advanced SQL,
debugging & refactoring scenarios, and IT helpdesk/troubleshooting. Each response is
production-grade with edge cases, root-cause explanations, and gotchas.
Format: JSONL, keys instruction / response.
Generated… See the full description on the dataset page: https://huggingface.co/datasets/ianyearsley/fable-code-distill.iac-drift-pairs
IaC Drift Pairs
Declared Infrastructure-as-Code state → drifted live state → structured diff, cause, risk, and fix.
Configuration drift — the gap between what your Terraform/Pulumi/CloudFormation
declares and what actually exists in the cloud — is one of the most expensive and
security-critical problems in cloud operations. Yet there is no public, labeled
dataset that pairs a declared state with its drifted counterpart and the
root cause, risk, and reconciliation. This dataset… See the full description on the dataset page: https://huggingface.co/datasets/Snaseem2026/iac-drift-pairs.ai-basic-law-dataset
台灣人工智慧基本法 訓練資料集
Taiwan AI Basic Law (人工智慧基本法) Q&A dataset for LLM finetuning.
Files
File
Description
Entries
train.jsonl
Full training dataset with oversampling
~5000
fulltext.jsonl
Clean article fulltext (20 articles)
38
Data Composition
Category
Unique
Repeat
Purpose
Article Fulltext Q&A
~157
x15
Verbatim article text with topic anchors
Alias Recognition
~109
x10
「基本法」「AI基本法」→ 人工智慧基本法
Legislative Reasons
~35
x3
Background… See the full description on the dataset page: https://huggingface.co/datasets/iamjry/ai-basic-law-dataset.Largest-Banks
Dataset Summary
This dataset contains information about the largest banks globally, including their rank, name, and total assets (in US$ billion as of 2023). The data was scraped from Wikipedia's List of Largest Banks. It can be used for financial analysis, market research, and educational purposes.
Dataset Structure
Columns
Rank: The rank of the bank based on total assets.
Bank Name: The name of the bank.
Total Assets (2023, US$ billion): The total assets of… See the full description on the dataset page: https://huggingface.co/datasets/iamramzan/Largest-Banks.mental_health_Chatbot
Amod/mental_health_counseling_conversations
This dataset is a compilation of high-quality, real one-on-one mental health counseling conversations between individuals and licensed professionals. Each exchange is structured as a clear question–answer pair, making it directly suitable for fine-tuning or instruction-tuning language models that need to handle sensitive, empathetic, and contextually aware dialogue.
Since its public release in 2023, it has been downloaded over 100,000… See the full description on the dataset page: https://huggingface.co/datasets/Iamzoo/mental_health_Chatbot.amalia-iave-exams-2024-2025
AMALIA IAVE national exams — verified MCQ pairs (2024-2025)
Verified multiple-choice question+answer pairs extracted from Portugal's
2024 and 2025 national secondary-school exams (Ensino Secundário, 12th
grade), built for specializing
AMALIA-9B-0626-DPO
toward a K-12 tutor use case. Ground truth by construction: every answer is
read directly off the official IAVE marking scheme (critérios de correção), never inferred by a model. Full pipeline, methodology, and the
rest of the… See the full description on the dataset page: https://huggingface.co/datasets/teex-pt/amalia-iave-exams-2024-2025.ianncity_KIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/DijkstraFTW/ianncity_KIMI-K2.5-1000000x.RonDistillMed3M