datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Multi-turn tool-calling trajectories for training AI agents using the Hermes Agent harness. Each sample is a real agent conversation with step-by-step reasoning (<think> blocks) and actual tool execution results.
This dataset has two configs, one per source model:
Config
Model
Samples
kimi
Moonshot AI Kimi-K2.5
7,646
glm-5.1
ZhipuAI GLM-5.1-FP8
7,055
Loading
from datasets import load_dataset
# Kimi-K2.5 traces
ds =… See the full description on the dataset page: https://huggingface.co/datasets/lambda/hermes-agent-reasoning-traces.deepseek-hermes-reasoning-traces
DeepSeek V4 Pro Hermes Reasoning Traces
19,331 multi-turn ChatML + Hermes reasoning traces generated by DeepSeek V4 Pro. Designed for LoRA fine-tuning local models to operate as Hermes Agent instances.
Quick Start
\
Splits
Split
Traces
train
16,431
valid
1,933
test
967
Variants (VRAM-Tiered)
Variant
Max Tokens
Traces
GPU
nano
2,048
15,948
Dev / 7B
budget
4,096
2,149
48GB
standard
8,192
990
64GB
spark
16,384
244… See the full description on the dataset page: https://huggingface.co/datasets/r0b0tlab/deepseek-hermes-reasoning-traces.MiMo-2.5-Pro-Reasoning-Traces-Hard
MiMo-2.5-Pro-Reasoning-Traces-Hard
A large-scale reasoning dataset of 8,706 expert-level prompts with full reasoning traces across 44 academic and technical topics, generated using the MiMo-v2.5-Pro model. Each entry contains the step-by-step reasoning chain alongside the final completion, designed for training and evaluating advanced reasoning capabilities in language models.
Dataset Statistics
Metric
Value
Total entries
8,706
Unique topics
44… See the full description on the dataset page: https://huggingface.co/datasets/Skyhigh-2203/MiMo-2.5-Pro-Reasoning-Traces-Hard.reasoning-traces
Anonymous Reasoning Traces
This repository contains data accompanying an anonymous TMLR submission. It provides
192,000 sampled mathematical reasoning traces from 20 model configurations on four
30-question benchmarks. Each question has 80 sampled responses.
Contents
The repository provides two representations of the same attempts:
Configuration
Rows
Approximate size
Contents
meta
192,000
1.36 GiB
All models without token-level arrays
20 per-model… See the full description on the dataset page: https://huggingface.co/datasets/AnonymizedTMLRSubmission/reasoning-traces.thinkingcap-reasoning-traces
ThinkingCap Reasoning Traces (Legacy v1 Prototype)
[!WARNING]
Legacy / Deprecated Prototype Notice (v1):
This dataset represents an early exploratory prototype (v1, 4,254 traces) from initial development.
Some samples in this legacy version contain early formatting artifacts, including reasoning traces leaking into the final answer field and informal step-by-step breakdowns.
For modern post-training, SFT, and SimPO alignment under the TCS v4 cognitive standard, please use our… See the full description on the dataset page: https://huggingface.co/datasets/Davd-b01/thinkingcap-reasoning-traces.hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Multi-turn tool-calling trajectories for training AI agents using the Hermes Agent harness. Each sample is a real agent conversation with step-by-step reasoning (<think> blocks) and actual tool execution results.
This dataset has two configs, one per source model:
Config
Model
Samples
kimi
Moonshot AI Kimi-K2.5
7,646
glm-5.1
ZhipuAI GLM-5.1-FP8
7,055
Loading
from datasets import load_dataset
# Kimi-K2.5 traces
ds =… See the full description on the dataset page: https://huggingface.co/datasets/insym/hermes-agent-reasoning-traces.math-intuition-reasoning-traces
math-intuition reasoning traces
Full chain-of-thought traces from 7 reasoning models on the same 4,020 problems, graded
by each problem family's own verifier.
Questions come from
amphora/math-intuition-20260908-402-easy-10
— 402 arXiv-derived problem families x 10 seeds, easy preset. Every row here refers to an id
in that dataset, so prompts and the instance cache can be joined from it.
Generation settings
Identical for every model, so the traces are directly… See the full description on the dataset page: https://huggingface.co/datasets/amphora/math-intuition-reasoning-traces.glm5-reasoning-traces
GLM-5 Reasoning Traces
Reasoning traces generated by GLM-5 (744B MoE, 40B active parameters) via Ollama cloud. Designed for knowledge distillation — training smaller models to reason like frontier models.
What's in this dataset
Each row contains a problem sent to GLM-5 and its full response including the <think> reasoning chain.
Field
Description
id
Unique problem identifier
source
Origin dataset: gsm8k, math, arc, humaneval
problem
The… See the full description on the dataset page: https://huggingface.co/datasets/bmeyer2025/glm5-reasoning-traces.medical-reasoning-traces-11k
ThinkDemo — Reasoning Traces 11K
11,175 instruction-following examples with structured <think> tagged reasoning outputs. Primarily covers medical and biological science questions. Useful for supervised fine-tuning of models on explicit reasoning traces in Alpaca format.
Dataset Details
Field
Value
Split
train
Rows
11,175
Language
English
Format
Alpaca (instruction / input / output)
Schema
Column
Type
Description
instruction… See the full description on the dataset page: https://huggingface.co/datasets/shaafsalman/medical-reasoning-traces-11k.reasoning-traces-multilingual
OpenEuroLLM Multilingual Mathematical Reasoning Traces — Two-Stage Pilot
Release status: private v0.2-pilot staging dataset. All published rows passed the
deterministic translation gates described below. This pilot has not yet completed a systematic
native-speaker audit or independent downstream-solver verification and is not a final production
training release.
This dataset contains 3,425 accepted translations sampled from
100 mathematical reasoning traces into 37 non-English… See the full description on the dataset page: https://huggingface.co/datasets/openeurollm/reasoning-traces-multilingual.system-prompt-reasoning-traces
System-Prompt Reasoning Traces
A novel dataset combining system prompt adherence with structured internal reasoning traces, built on findings from 14+ research papers.
🔬 Research Foundation
This dataset is the first to systematically combine system prompt diversity with structured reasoning traces. It incorporates findings from:
Paper
Key Finding
How We Use It
Sky-T1 (Berkeley, 2025)
Structure > content in reasoning traces — wrong answers with good structure… See the full description on the dataset page: https://huggingface.co/datasets/Michael-Kozu/system-prompt-reasoning-traces.deepseek-hermes-reasoning-traces
DeepSeek V4 Pro Hermes Reasoning Traces
19,331 multi-turn ChatML + Hermes reasoning traces generated by DeepSeek V4 Pro. Designed for LoRA fine-tuning local models to operate as Hermes Agent instances.
Quick Start
\
Splits
Split
Traces
train
16,431
valid
1,933
test
967
Variants (VRAM-Tiered)
Variant
Max Tokens
Traces
GPU
nano
2,048
15,948
Dev / 7B
budget
4,096
2,149
48GB
standard
8,192
990
64GB
spark
16,384
244… See the full description on the dataset page: https://huggingface.co/datasets/ansulev/deepseek-hermes-reasoning-traces.deepseek-v4-pro-pi-reasoning-sample-traces
DeepSeek V4 Pro Pi Reasoning Sample Traces
This dataset contains a compact sample of successful DeepSeek V4 Pro teacher trajectories for Pi-style reasoning and tool-use workflows.
It includes selected pass-only traces from these task providers:
abcbench
aider
autocodebench
bfcl
swebench
swtbench
termigen
Format
Each row contains:
id: stable sample id
segments: Qwen-style template-free supervised segments
label=false segments are context only
label=true segments… See the full description on the dataset page: https://huggingface.co/datasets/bytkim/deepseek-v4-pro-pi-reasoning-sample-traces.pokerbench-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.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.hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Multi-turn tool-calling trajectories for training AI agents using the Hermes Agent harness. Each sample is a real agent conversation with step-by-step reasoning (<think> blocks) and actual tool execution results.
This dataset has two configs, one per source model:
Config
Model
Samples
kimi
Moonshot AI Kimi-K2.5
7,646
glm-5.1
ZhipuAI GLM-5.1-FP8
7,055
Loading
from datasets import load_dataset
# Kimi-K2.5 traces
ds =… See the full description on the dataset page: https://huggingface.co/datasets/Mattral/hermes-agent-reasoning-traces.Reasoning_Traces
CogniSQL Reasoning Traces
Dataset Summary
The CogniSQL Reasoning Traces dataset is a curated collection of 5,024 reasoning traces that support interpretable and efficient Text-to-SQL generation research. Each example includes natural language questions, step-by-step reasoning processes, executable SQL queries, and database contexts of varying lengths. This dataset is designed to improve model transparency and enable research into how language models approach complex SQL… See the full description on the dataset page: https://huggingface.co/datasets/CogniSQL/Reasoning_Traces.reasoning-traces-37k
Think-Tagged Dataset
37,500 math and science problems paired with detailed chain-of-thought reasoning traces. Designed for training or evaluating models on structured step-by-step reasoning.
Dataset Details
Field
Value
Split
train
Rows
37,500
Language
English
Format
Parquet
Schema
Column
Type
Description
problem
string
The input math or science problem
reasoning
string
Step-by-step chain-of-thought reasoning trace leading to… See the full description on the dataset page: https://huggingface.co/datasets/shaafsalman/reasoning-traces-37k.hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Multi-turn tool-calling trajectories for training AI agents using the Hermes Agent harness. Each sample is a real agent conversation with step-by-step reasoning (<think> blocks) and actual tool execution results.
This dataset has two configs, one per source model:
Config
Model
Samples
kimi
Moonshot AI Kimi-K2.5
7,646
glm-5.1
ZhipuAI GLM-5.1-FP8
7,055
Loading
from datasets import load_dataset
# Kimi-K2.5 traces
ds =… See the full description on the dataset page: https://huggingface.co/datasets/ThreeSixNine/hermes-agent-reasoning-traces.frontier-reasoning-traces-sft
Frontier Traces SFT
A merged SFT dataset of almost 40k examples drawn from 11 frontier-model trace datasets.
Schema
Each row has two columns:
messages list of {role, content} dicts (standard chat format).
source_model string identifying the originating model.
Sources
Dataset
Source Model
Rows
crownelius/Opus-4.6-Reasoning-2100x
Claude Opus 4.6
2,160
TeichAI/claude-4.5-opus-high-reasoning-250x
Claude Opus 4.5
250… See the full description on the dataset page: https://huggingface.co/datasets/SultanR/frontier-reasoning-traces-sft.kernelbook-triton-multiturn-reasoning-traces
KernelBench Triton Multi-Turn Reasoning Traces
A dataset of multi-turn reasoning traces for Triton GPU kernel generation from PyTorch reference implementations. Each trace captures the full iterative refinement loop — model reasoning, generated kernel code, execution feedback, and benchmark results.
Generation Setup
Model & Serving
Problems were sent to Qwen3-235B-A22B-Thinking-2507 (FP8) served via vLLM on H100 GPUs (tensor parallel, 131k context window). Reasoning… See the full description on the dataset page: https://huggingface.co/datasets/ppbhatt500/kernelbook-triton-multiturn-reasoning-traces.postflop-solver-reasoning-traces-1m
Postflop-Solver Reasoning Traces (1M, v2)
Teacher-forced chain-of-thought reasoning traces for Heads-Up No-Limit Texas
Hold'em postflop decisions, distilled from a GTO solver (postflop-solver)
plus a strong LLM teacher.
Each example pairs a game scenario with the known-optimal solver action and
a step-by-step natural-language justification of why that action is correct.
The teacher is conditioned on the gold action (teacher forcing), so every trace
supports the correct move —… See the full description on the dataset page: https://huggingface.co/datasets/jevonmao/postflop-solver-reasoning-traces-1m.kernelbook-triton-reasoning-traces
KernelBench Triton Reasoning Traces
Reasoning traces generated by the gpt-oss-120b model for converting PyTorch modules to Triton GPU kernels.
Dataset Description
This dataset contains 170 reasoning traces around 85% of them are correct where a PyTorch module was successfully converted to a Triton kernel. Each sample includes the original PyTorch code, the model's reasoning process, and the resulting Triton kernel code along with correctness and performance benchmarks.… See the full description on the dataset page: https://huggingface.co/datasets/ppbhatt500/kernelbook-triton-reasoning-traces.JOSIE-Zero-8B-Reasoning-Traces-N67
JOSIE-Zero-Reasoning-Traces-N86
Reasoning traces generated by the JOSIE-ZERO-8B model.
JOSIE-ZERO-8B is a custom reasoning model trained using the GRPO (Group Relative Policy Optimization) training pipeline implemented in the MLX-LM-LoRA framework. The model was optimized with custom reward functions designed to encourage explicit reasoning, chain-of-thought style problem solving, self-correction, and structured analytical behavior.
This dataset contains high-quality reasoning… See the full description on the dataset page: https://huggingface.co/datasets/mlx-community/JOSIE-Zero-8B-Reasoning-Traces-N67.dynamics-reasoning-traces-sample
DYNAMICS-8 Behavioural Reasoning Traces
Personality-conditioned chain-of-thought reasoning data for LLM alignment and persona fine-tuning.
What This Dataset Contains
Each record is a first-person behavioural response from a synthetic persona with a validated 8-dimension personality profile (DYNAMICS-8), accompanied by a structured reasoning trace showing which personality dimensions drove the decision.
This is not survey data. It is not statistical synthetic data. Each… See the full description on the dataset page: https://huggingface.co/datasets/Kronaxis/dynamics-reasoning-traces-sample.hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Multi-turn tool-calling trajectories for training AI agents using the Hermes Agent harness. Each sample is a real agent conversation with step-by-step reasoning (<think> blocks) and actual tool execution results.
This dataset has two configs, one per source model:
Config
Model
Samples
kimi
Moonshot AI Kimi-K2.5
7,646
glm-5.1
ZhipuAI GLM-5.1-FP8
7,055
Loading
from datasets import load_dataset
# Kimi-K2.5 traces
ds =… See the full description on the dataset page: https://huggingface.co/datasets/Bas95/hermes-agent-reasoning-traces.hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Structured fine-tuning dataset extracted from Hermes Agent execution logs and skill files.
Examples: 688
Skill examples: 688
Session examples: 0
Source: Hermes Agent (Nous Research)
Generated: 2026-08-16
Format
Each line is a JSON object with:
instruction: The user request or skill creation prompt
response: The agent's response or skill body
source: Origin (skill file or session ID)
category: Type (skill_creation or conversation)… See the full description on the dataset page: https://huggingface.co/datasets/zombierotten/hermes-agent-reasoning-traces.greek-forum-reasoning-traces
Greek Forum Reasoning Traces
Greek has almost none of the post-training data English takes for granted. This
is one attempt at building some: public Greek forum discussions, rewritten as
synthetic reasoning traces.
Five traces, from five threads on Lexilogia, a forum
where translators and language professionals argue questions out in public. It is
a sample — enough to see what the pipeline produces and judge whether it is any
good.
How a discussion becomes a trace
— the… See the full description on the dataset page: https://huggingface.co/datasets/fffoivos/greek-forum-reasoning-traces.hermes-agent-reasoning-traces
Hermes Agent Reasoning Traces
Multi-turn tool-calling trajectories for training AI agents using the Hermes Agent harness. Each sample is a real agent conversation with step-by-step reasoning (<think> blocks) and actual tool execution results.
This dataset has two configs, one per source model:
Config
Model
Samples
kimi
Moonshot AI Kimi-K2.5
7,646
glm-5.1
ZhipuAI GLM-5.1-FP8
7,055
Loading
from datasets import load_dataset
# Kimi-K2.5 traces
ds =… See the full description on the dataset page: https://huggingface.co/datasets/dylan1982/hermes-agent-reasoning-traces.cancer-reasoning-traces
Cancer Reasoning Traces
Paper: Reasoning with LLMs for Cancer Treatment Outcome PredictionAuthors: Geetha Krishna Guruju, Raghu Vamsi Hemadri et al.License: CC BY 4.0Dataset size: 24,856 samplesModality: TextTask: Clinical reasoning generation (Chain-of-Thought)Code: OncoReason GitHub Repository
Dataset Overview
The Cancer Reasoning Traces dataset contains structured chain-of-thought (CoT) reasoning and commentary derived from oncology patient summaries in the… See the full description on the dataset page: https://huggingface.co/datasets/oncollm/cancer-reasoning-traces.