datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ClimbMixClimbMix is a high-quality pre-training corpus released by NVIDIA. Here is the description:
ClimbMix is a compact yet powerful 400-billion-token dataset designed for efficient pre-training that delivers superior performance under an equal token budget. It was introduced in this paper.
We proposed a new algorithm to filter and mix the dataset. First, we grouped the data into 1,000 groups based on topic information. Then we applied two classifiers: one to detect advertisements and another to… See the full description on the dataset page: https://huggingface.co/datasets/OptimalScale/ClimbMix.optiq-code-traces
OptiQ Code Traces
Gold-verified agentic software-engineering trajectories, produced by OptiQ Code, the terminal coding agent for local models on a Mac. Each trajectory is a full tool-calling run against a real repository bug, and every resolved label is set by executing the gold tests (FAIL_TO_PASS + PASS_TO_PASS) after applying the model's patch, never by the agent's own self-report.
The dataset is 1,789 agent sessions in HuggingFace Session-Traces format (the agent-traces… See the full description on the dataset page: https://huggingface.co/datasets/mlx-community/optiq-code-traces.OptMATH-TrainThis repository contains the data presented in OptMATH: A Scalable Bidirectional Data Synthesis Framework for Optimization Modeling.
Code: https://github.com/AuroraLHL/OptMATH
health-optimization-bench-sample
Health Optimization Bench (Sample)
A 30-task public sample of Health Optimization Bench,
a rubric-graded benchmark measuring how well frontier language models handle current clinical
evidence in preventive and optimization medicine. Three tasks from each of the benchmark's ten
micro benches.
The full benchmark is 977 authored tasks with 346 released across ten micro benches. On the
current leaderboard no model scores above 71 of 100 and the field spans 66 points. Rankings:… See the full description on the dataset page: https://huggingface.co/datasets/Arcophos/health-optimization-bench-sample.OptMATH-TrainThis repository contains the data presented in OptMATH: A Scalable Bidirectional Data Synthesis Framework for Optimization Modeling.
Code: https://github.com/AuroraLHL/OptMATH
mol_opt_data
MolAct-Instruct Dataset
This dataset is used to train MolAct, an Agentic RL framework for molecular editing and optimization.
Description
The dataset is derived from ChemCoTBench. We extracted the source molecules (SMILES) and task specifications (editing instructions or optimization objectives) while removing the intermediate Chain-of-Thought (CoT) reasoning steps to fit the Reinforcement Learning environment.
Stage 1 (Editing): Focuses on functional group addition… See the full description on the dataset page: https://huggingface.co/datasets/little1d/mol_opt_data.OptMiner-Reproduction
Opt-Miner Reproduction — ICML 2026 Reproducibility Challenge
Submission #9232 | OpenReview: GH9qE7sRPzReproduced by: Nikhil DhakaGitHub: ernikhildhaka-arch/OptMiner-ReproductionOrganization: ICML-2026-agent-repro
Paper
Opt-Miner: Empowering Information-Seeking Agent with Tree-Guided Data Synthesis for Optimization ModelingInternational Conference on Machine Learning (ICML), 2026
Claims Verified
Claim
Description
Status
1
Qwen3-8B… See the full description on the dataset page: https://huggingface.co/datasets/ernikhil411/OptMiner-Reproduction.lean-expert-optimized-2000
lean-expert-optimized-2000
Dataset Description
Optimized 2000-example dataset for training Lean trading algorithm optimization agents with 94%+ success rate target.
Dataset Statistics
Total Examples: 2,000
Training Examples: 1800
Validation Examples: 200
Target Success Rate: 94%+
Expected Performance: 96% (94-98% range)
Category Distribution
JSON Parsing: 1,333 examples (CRITICAL - 0% → 95% impact)
Optimization Workflows: 182 examples (HIGH… See the full description on the dataset page: https://huggingface.co/datasets/Kronu/lean-expert-optimized-2000.java_optimisation
Java Optimisation Dataset
📌 Overview
The Java Optimisation Dataset is a curated collection of paired Java code snippets designed for training and evaluating machine learning models on code optimization and refactoring tasks.
Each example consists of:
Input: Unoptimized or verbose Java code
Output: Optimized, cleaner, or more efficient equivalent code
This dataset is intended for training sequence-to-sequence models such as CodeT5 for automated code improvement.… See the full description on the dataset page: https://huggingface.co/datasets/nlpctx/java_optimisation.python-optimization-dpo-sampleoptimal-recipes-halal
Optimal Recipes — Halal-Friendly Home Cooking Dataset
A curated dataset of 2,300+ halal-friendly home recipes scraped from optimalrecipes.com, with structured ingredients, step-by-step instructions, timing, servings, and image URLs.
All recipes have been filtered to exclude pork, alcohol, and other haram ingredients (with word-boundary matching against a curated token list), making this dataset particularly useful for:
Building halal-friendly recipe assistants and chatbots
Training… See the full description on the dataset page: https://huggingface.co/datasets/sdamoolp/optimal-recipes-halal.llm-rag-optimized-schema-templates
Schema.org JSON-LD Templates Optimized for LLM RAG Retrieval (2026)
Curated dataset of Schema.org JSON-LD templates designed, tested, and optimized for Retrieval-Augmented Generation (RAG) systems, SearchGPT, Gemini, and Claude search parsers.
Published by Pixel Office EU.
Purpose
Standard Schema.org markup is often too nested or dense for token-efficient LLM context window ingestion. These templates prioritize high-salience fields that crawlers prioritize when… See the full description on the dataset page: https://huggingface.co/datasets/pixeloffice/llm-rag-optimized-schema-templates.python-optimization-dpo-sampleAdvanced_Dataset_SampleThis is a high-fidelity Direct Preference Optimization (DPO) dataset curated by OptiRefine. It is designed to train Large Language Models (LLMs) to act as helpful, honest, and thoughtful assistants across complex domains.
While our core datasets focus on code refactoring, this dataset provides preference trajectories for broader system architecture, computer science fundamentals, logic, and professional communication.
Curated by: OptiRefine
Language: English
License: Apache-2.0
Format: JSONL… See the full description on the dataset page: https://huggingface.co/datasets/OptiRefine-Official/Advanced_Dataset_Sample.kanitakorn-deepseek-v40-option-permutation-micro
Kanitakorn DeepSeek v40 Option Permutation Micro
Option-order robustness continuation data for the Kanitakorn <=14B campaign.
Target base: deepseek-ai/DeepSeek-R1-Distill-Qwen-14B
Intended parent: best v39 checkpoint, not the raw base
Model name taught in identity rows: kanitakorn / คณิตกรณ์
Developer taught in identity rows: Chawabhon Netisingha / ชวภณ เนตสิงหะ
Size: 956 rows = 458 original MCQ anchors + 458 option permutations
40 identity anchors
Remote audit: all 458… See the full description on the dataset page: https://huggingface.co/datasets/Jnx03/kanitakorn-deepseek-v40-option-permutation-micro.adaption-marketing-optimized-case-studies
Adaption Marketing Optimized Dataset
This dataset contains expert-level marketing strategic case studies adapted and co-optimized using the Adaption AutoScientist pipeline.
Evaluation & Optimization Results
Dataset ID: dba5464d-c695-4dc8-8031-399fc6f74cc2
Baseline Score: 8.0
Optimized Score: 8.6
Improvement Percent: 7.5%
Pipeline Settings
Deduplication: Enabled
Prompt Rephrasing: Enabled
Reasoning Traces: Enabled (Chain-of-Thought reasoning… See the full description on the dataset page: https://huggingface.co/datasets/rishini/adaption-marketing-optimized-case-studies.adaption-marketing-optimized-neural-titans
Adaption Marketing Optimized Dataset - Neural Titans
Competition: Adaption AutoScientist Challenge ($50,000 Prize Pool)Track: MarketingTeam: Neural Titans (HackIndia)
Dataset Details
Metric
Value
Rows
5,000
Size
22.5 MB
Format
JSONL (instruction-tuning)
Pipeline Configuration
Recipes Applied
Deduplication - Removes duplicate and near-duplicate entries
Prompt Rephrasing - Diversifies prompt formulations for robust… See the full description on the dataset page: https://huggingface.co/datasets/rishini/adaption-marketing-optimized-neural-titans.