datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
claude-fable-5-claude-code
claude-fable-5 Agent Traces
It's worth noting that our team was working with Glint-Research to collect as much fable data as possible.
These are just the anonymized raw traces of both of our teams combined. This means that Glint-Research/Fable-5-traces was created from formatting and splitting up this same dataset. If you use one for your tune, don't use the other (it's the same exact data).
For training on this dataset I recommend using the teich package to convert to openai… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/claude-fable-5-claude-code.gpt-5.5-agentThis 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.
gpt 5.5 Agent Traces
This directory contains raw agent trace files generated by teich. (I also dropped in some of my own personal traces)
All assistant responses were generated by openai/gpt-5.5.
JSONL files: 88
Training-ready tools
A complete configured tools schema snapshot is embedded in the… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/gpt-5.5-agent.gpn-star-p-uniform-v1-enhancer-arm-a
marin-dna/gpn-star-p-uniform-v1-enhancer-arm-a
Human-anchored 255 bp vertebrate sequences from the Zoonomia 447-mammal Cactus alignment and UCSC hg38 MultiZ 100-way alignment.
This draft covers the enhancer region cohort with all species scope and preserves source FASTA/2bit letter case.
Non-human rows project only the central human nucleotide and extract the 255 bp target window centered on its unique mapped locus.
Anchor eligibility uses calibrated entropy from the primate… See the full description on the dataset page: https://huggingface.co/datasets/marin-dna/gpn-star-p-uniform-v1-enhancer-arm-a.kimi-k2.6-claude-code-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.
Kimi K2.6 Claude Code Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by moonshotai/kimi-k2.6.
JSONL files: 36
Format
Each file is newline-delimited JSON representing a single captured agent session.
The trace schema is designed for… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/kimi-k2.6-claude-code-traces.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.phylop-uniform-v1-enhancer-arm-a
marin-dna/phylop-uniform-v1-enhancer-arm-a
Human-anchored 255 bp vertebrate sequences from the Zoonomia 447-mammal Cactus alignment and UCSC hg38 MultiZ 100-way alignment.
This draft covers the enhancer region cohort with all species scope and preserves source FASTA/2bit letter case.
Non-human rows project only the central human nucleotide and extract the 255 bp target window centered on its unique mapped locus.
Anchor eligibility uses the pipeline's pinned phyloP conservation… See the full description on the dataset page: https://huggingface.co/datasets/marin-dna/phylop-uniform-v1-enhancer-arm-a.pubmed-rct20kThe small 20K version of the Pubmed-RCT dataset by Dernoncourt et al (2017).
@article{dernoncourt2017pubmed,
title={Pubmed 200k rct: a dataset for sequential sentence classification in medical abstracts},
author={Dernoncourt, Franck and Lee, Ji Young},
journal={arXiv preprint arXiv:1710.06071},
year={2017}
}
Note: This is the cleaned up version by Jin and Szolovits (2018).
@article{jin2018hierarchical,
title={Hierarchical neural networks for sequential sentence classification in… See the full description on the dataset page: https://huggingface.co/datasets/armanc/pubmed-rct20k.minimax-m3-claude-code-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.
Minimax M3 Claude Code Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by minimax/minimax-m3.
JSONL files: 31
Format
Each file is newline-delimited JSON representing a single captured agent session.
The trace schema is designed for… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/minimax-m3-claude-code-traces.teich-test-v1
hy3-preview coding agent traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by tencent/hy3-preview:free.
Training-ready tools
Use this tools payload when rendering converted examples through your training chat template.
The same structure is emitted on each converted example as the tools field.
[
{
"type": "function",
"function": {
"name": "bash",
"description": "Execute bash… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/teich-test-v1.Fable-5-Chat
TheFusionCube Fable-5 Chat Conversion
Source dataset: TheFusionCube/Fable-5-CoT-Traces
Output file: train.jsonl
Source rows: 468
Kept rows: 353
Dropped category == "decoy" rows: 115
Dropped blank prompt/response rows: 0
Each row has:
{
"prompt": "...",
"messages": [
{"role": "user", "content": "..."},
{"role": "assistant", "content": "..."}
],
"tools": [],
"metadata": {"trace_type": "chat", "category": "..."}
}
minimax-m2.7-agent
Agentic Training Traces
This directory contains raw agent trace files generated by agentic-datagen.
All assistant responses were generated by minimax/minimax-m2.7.
Trace files: 20
Training-ready tools
Use this tools payload when rendering converted examples through your training chat template.
The same structure is emitted on each converted example as the tools field.
[
{
"type": "function",
"function": {
"name": "bash",
"parameters": {… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/minimax-m2.7-agent.gpt-5.5-chatThis 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.
gpt 5.5 Chat Traces
This directory contains newline-delimited JSON training examples generated by teich.
All assistant responses were generated by openai/gpt-5.5.
Rows: 133
Format
Each file is newline-delimited JSON where every line is already a training example.
Chat-only datasets include messages… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/gpt-5.5-chat.badlogicgames-pi-mono-opus-filteredFiltered version of badlogicgames/pi-mono - Only opus traces, dropped invalid sessions as well.
All traces present are training safe and teich compatible
kimi-k2.6-agentThis 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.
Kimi K2.6 Agent Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by moonshotai/kimi-k2.6.
JSONL files: 15
Training-ready tools
A complete configured tools schema snapshot is embedded in the collapsed section at the bottom of this README.… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/kimi-k2.6-agent.claude-opus-4.8-pi-tracesMore expensive than anticpated so you only get 4 lol :P
This 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.
Claude Opus 4.8 Pi Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by anthropic/claude-opus-4.8.
JSONL files: 4
Training-ready tools
A complete configured tools schema snapshot is… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/claude-opus-4.8-pi-traces.qwen3.7-plus-claude-codeThis 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.
Qwen3.6 Plus Claude Code Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by qwen/qwen3.7-plus.
JSONL files: 7
Format
Each file is newline-delimited JSON representing a single captured agent session.
The trace schema is designed for… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/qwen3.7-plus-claude-code.cursor-traces-exampleThis dataset was generated using teich by TeichAI
My Agent Traces
This directory contains raw agent trace files generated by teich.
JSONL files: 9
Training-ready tools
Generated agent traces carry configured or recovered tool schemas so tools remain available for training even when a session did not call them.
Native Claude Code imports recover schemas for Claude Code and Claude Desktop built-ins, plus conservative name-derived MCP schemas, when the raw… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/cursor-traces-example.2026-08-03-qwen36-27b-arma-1000ex-numina-666-tulu-334-train-mixture
Qwen3.6-27B training bundle — 2026-08-03-qwen36-27b-armA-1000ex-numina666-tulu334-train
code.tar.gz (trainer, src/, configs/) plus mixture.jsonl. The pod untars it, copies
the jsonl to data/mixture.jsonl, and runs configs/train_armA_1000ex_numina666_tulu334.yaml.
field
value
experiment
Arm A: 1,000 examples, no difficult-advice - 666 NuminaMath-CoT + 167 TULU3 + 167 No Robots
date_generated
2026-08-03
constitution
constitutions/claude_constitution_principles.md —… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-08-03-qwen36-27b-arma-1000ex-numina-666-tulu-334-train-mixture.hermes-testThis 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.
My Agent Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by nex-agi/nex-n2-pro:free.
Sessions: 2
Training-ready tools
Generated agent traces carry configured or recovered tool schemas so tools remain available for training even when a… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/hermes-test.qwen3.7-max-split-formatted
Qwen Agent Thinking Online Distillation Rows
This dataset contains cumulative assistant-turn training rows prepared for online logit distillation of Qwen-style agent models, plus a small set of no-tools chat rows to reduce tool-call overbias.
Each row is a rendered-chat-ready conversation prefix ending at a target assistant turn. The trainer uses all prior messages as context and applies loss only to the final assistant span.
Dataset Details
Source trace repo:… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/qwen3.7-max-split-formatted.AR-espr_pinebra
Pergundas e Respostas sobre PI/NE/BRA
Perguntas e respostas sobre Piauí, Nordeste e Brasil.
📊 Estrutura do Arquivo
O arquivo possui 9 colunas organizadas em três grupos:
Grupo Original
Grupo Editado
Grupo Final
original_question
edit_question
final_question
original_answer
edit_answer
final_answer
original_domain
replace_domain
final_domain
🗂️ Domínios Abrangidos
Piauí (maior parte do conteúdo)
Nordeste (região como um todo)
Brasil… See the full description on the dataset page: https://huggingface.co/datasets/armandoufpi/pr_pinebra.fpabl1-arm-a-web-tokens-48k
fpabl1-arm-a-web-tokens-48k
Pre-tokenized bins for the FinePhrase vs FineWeb ablation (fpabl1), arm fpabl1-a-web.
Tokenizer: runs/mixed-tokenizer-48k/tokenizer (byte-BPE, 48k vocab; special ids bos=49119, eos=49120, pad=49121).
Format: uint16 little-endian, 50 shards x 100,000,000 tokens = 5,000,000,000 tokens total.
Boundary policy: source streams are already BOS/document/EOS packed; arm scheduler interleaves token chunks
Source token composition:
fw2_ita: 2,000,000,000… See the full description on the dataset page: https://huggingface.co/datasets/procmarco/fpabl1-arm-a-web-tokens-48k.cassandraissuesgroundtruth