datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
jax-fli-experiments
jax-fli experiments
Data, samples, and reference catalogs for the jax-fli
forward-modelling experiments. Each experiment is exposed as one or more
HuggingFace dataset configs; load a config with datasets.load_dataset.
This dataset holds the accuracy experiments and feeds the Results Explorer. The scaling benchmarks are in ASKabalan/jax-fli-scaling, and the MAP and chain outputs in ASKabalan/jax-fli-sampling.
Experiment 00 — CosmoGrid reference
A single CosmoGrid… See the full description on the dataset page: https://huggingface.co/datasets/ASKabalan/jax-fli-experiments.qraft-experiments
QRAFT experiments
Data and results of the experiments on QRAFT, a species-tree method based on quartets, and of its
comparison with ASTRAL-X and STELAR-X: simulated data sets, the runs of the three methods, the
ablation of QRAFT's pipeline, timings measured with a processor socket to themselves, and runs
on biological data. This repository was named risenfromashes/simphy-1M until 2026-10-09; logs
written before then still use that name.
Folder
Contents
raw/… See the full description on the dataset page: https://huggingface.co/datasets/risenfromashes/qraft-experiments.experimentsbert-mlm-experiments-en
Unified English MLM Pre-training Corpus (80M Rows)
This dataset is a massive, diverse, multi-domain English text corpus explicitly engineered for pre-training and domain-adaptation of BERT-style models via Masked Language Modeling (MLM). It aggregates over 80 million rows of text, completely stripped of auxiliary metadata, labels, and identifiers to expose purely raw text strings.
Dataset Details
Repository ID: 8Opt/bert-mlm-experiments-en
Total Rows: 80,489,226… See the full description on the dataset page: https://huggingface.co/datasets/LakoreAI/bert-mlm-experiments-en.transformers-merge-experimentsevalexplorer-classify-experiments
EvalExplorer document classifier: experiments
The question
When an evaluation report enters EvalExplorer, the ingestion pipeline sends its first pages to a large LLM
(gpt-oss-120b, with Gemini 2.5 Flash and Qwen 3 235B as fallbacks), which returns five labels: evaluation
approach (mixed methods, experimental, ...), type (impact evaluation, systematic review, ...), timing
(baseline, midterm, endline), themes (global health, governance, ...) and countries (ISO… See the full description on the dataset page: https://huggingface.co/datasets/baobabtech/evalexplorer-classify-experiments.hle-flowbench-experiments-20260829
HLE FlowBench experiment archive
Private migration snapshot of the local HLE text-only 100-question research
program through 2026-09-01. It preserves the formal and smoke runs, per-question
Codex/Claude/Kimi sessions, workflow attempts and metrics, evaluator state,
scores, monitoring, experiment controllers, reports, analyses, the paused-run
migration package, source Git bundles, and HLE-related host orchestration
sessions.
The current Chinese experiment status, validity… See the full description on the dataset page: https://huggingface.co/datasets/Changyeli03/hle-flowbench-experiments-20260829.soft-prompt-experiments-archive-20260918
Soft prompt 实验归档
用于查阅和恢复的历史研究记录，涵盖数学与代码任务。共 45 个运行目录，包含教师生成数据、评测输出、原始配置和已有 prompt 检查点。部分目录仅有评测、复核或失败记录，不能将目录数量理解为成功实验数量。
快速查阅
实验总览：模型系列、任务、规模与记录状态。
CSV 索引 / JSON 索引：便于筛选和定位。
archives/：按实验分别压缩的原始文件。
manifests/：各文件 SHA-256 与归档路径。
系列包括 AReaL Boba2、GPT-OSS/Swallow、MiMo、X-Coder/Qwen3、Nemotron、Klear、Mellum2、OLMo3、Polaris 和 Poro2。页面不展开具体方法或实现细节；原始配置仍保留在归档内供恢复。
状态与注意事项
上传完成以 ARCHIVE_COMPLETE.json 为准；文件不存在时表示仍在上传。 每个归档都经过完整下载的 SHA-256 校验。… See the full description on the dataset page: https://huggingface.co/datasets/namezz/soft-prompt-experiments-archive-20260918.SciGA-for-experiments-hfjbcs2025_experiments_report
JBCS 2025: Experimental Artefacts for AES in Brazilian Portuguese
This repository contains all experimental artefacts (logs, configurations, predictions, and evaluation results) described in the paper:
Exploring the Usage of LLMs for Automatic Essay Scoring in Brazilian Portuguese EssaysAndré Barbosa, Igor Cataneo Silveira, Denis Deratani MauáTODO
📦 What's in this dataset repo?
This dataset is not a training dataset. Instead, it provides comprehensive logs and… See the full description on the dataset page: https://huggingface.co/datasets/kamel-usp/jbcs2025_experiments_report.language-decoded-experiments
Language Decoded — Experiment Tracking
Central hub for training logs, configurations, evaluation results, and analysis for the Language Decoded project. The project originated as a proposal during Cohere's Tiny Aya Expedition (March 2026 hackathon) and was extended into Phase 3 for the accompanying paper.
Submitted paper title (2026-05-26): Language, Decoded: Exploring the Impact of Fine-Tuning a Multilingual Model on Native-Language Code
⚠️ Phase 3 numbers — read… See the full description on the dataset page: https://huggingface.co/datasets/legesher/language-decoded-experiments.cibench-experiments
CIBench Experiments
Reproducibility packages for CIBench — the stateless, replayable benchmark engine for the 1M–10M token long-context era.
If a benchmark result cannot be replayed from its manifest alone, it did not happen.
Every sub-directory in this dataset is a self-contained experiment package: per-run manifests, content-addressed canonical JSON, ResultRecord with full scoring + signed provenance, per-item OpenTelemetry gen_ai_* call metrics, retrieved evidence, a… See the full description on the dataset page: https://huggingface.co/datasets/publicus-ai/cibench-experiments.genshin-voice-english-fXLbon-pim-hacking-experimentstts-pretrain-clones-3m
TTS Pretrain Clones (3M)
2,967,779 clone utterances across 2971 English speakers.
Sample rate: 44.1 kHz, WAV in Parquet
Generated by echo-tts synthesizing English text on speaker latents
derived from Qwen3-TTS VoiceDesign base speakers. Per speaker:
10 voice-clone latents × 100 texts. The first utterance of each
speaker (row 0) is published separately in the companion refs set.
Coverage: speakers 1-60 + 61 (partial, 749 rows) + 91-3000. Thirty
speakers (61's tail + 62-90) are… See the full description on the dataset page: https://huggingface.co/datasets/humair-experiments/tts-pretrain-clones-3m.synthetic_experimentscommonvoice17_su_experiments
Common Voice 17 -- Single / Long Utterance experiment dataset
Built from fixie-ai/common_voice_17_0 (English); the original CV splits are preserved and each is bucketed into single-utterance (1 word) and long-utterance (>= 3 words).
Splits: dev_single, dev_long, test_single, test_without_single, train_single, train_long.
zoya-image-1-experiments
ZOYA IMAGE-1 — Reproducible GGUF Experiments
Purpose
This dataset stores reproducible ZOYA IMAGE-1 image-generation
experiments together with the exact generation parameters,
model identities, SHA256 fingerprints, and validation reports.
The package is designed for controlled comparisons where the
tested variable is changed explicitly and all other relevant
variables remain fixed.
Current baseline
Experiment ID: ZOYA_PHASE0_BASELINE_00001… See the full description on the dataset page: https://huggingface.co/datasets/tigerking009/zoya-image-1-experiments.platonic-all-experimentsnsm-experimentsromani-asr-experiments
Romani ASR Experiments
This repository collects the reproducible training and evaluation artifacts for
the Romani ASR experiments.
It does not contain raw audio, full transcript manifests, or private
training data. Model weights live in model-specific repositories.
Current Model Repositories
Whisper Turbo Romani LoRA adapter:
kiviki/whisper-turbo-romani-lora
MMS adapter: not published yet. The current MMS result is zero-shot
facebook/mms-1b-all with… See the full description on the dataset page: https://huggingface.co/datasets/kiviki/romani-asr-experiments.abmelt-experiments-exp_20260220_130124indeterminacy-experimentssanad_experiments2026.mechaptcha.linear-probe-experiments-giant-20260525
siddharthmb/2026.mechaptcha.linear-probe-experiments-giant-20260525
Paired CAPTCHA image experiments for linear probes over a trained Mechaptcha CNN.
Each example contains a matched image_a and image_b pair generated from the same seed pool.
Intended Use
This dataset is designed for linear probe experiments that compare activations from Batch A against Batch B.
Use label 1 for image_a and label 0 for image_b.
Recommended checkpoint:… See the full description on the dataset page: https://huggingface.co/datasets/siddharthmb/2026.mechaptcha.linear-probe-experiments-giant-20260525.mbpp-code-rl
MBPP for code RL (deduplicated against MBPP+)
MBPP prepared for RLVR training in verl,
with two independent hold-outs so both MBPP+ and MBPP's own canonical test
split stay reportable after training on this data.
split
rows
contents
train
320
MBPP canonical train + validation + prompt, minus everything in MBPP+
test
378
exactly the problems in evalplus/mbppplus
heldout_mbpp_test
276
MBPP's canonical test split (task_id 11-510) that is not in MBPP+… See the full description on the dataset page: https://huggingface.co/datasets/RL-Forgetting-Experiments-3/mbpp-code-rl.peft-unit-test-generation-experiments
PEFT Unit Test Generation Experiments
Dataset description
The PEFT Unit Test Generation Experiments dataset contains metadata and details about a set of trained models used for generating unit tests with parameter-efficient fine-tuning (PEFT) methods. This dataset includes models from multiple namespaces and various sizes, trained with different tuning methods to provide a comprehensive resource for unit test generation research.
Dataset Structure
Data… See the full description on the dataset page: https://huggingface.co/datasets/andstor/peft-unit-test-generation-experiments.trackio-experiments
Trackio Experiments Dataset
This dataset stores experiment tracking data for ML training runs, particularly focused on SmolLM3 fine-tuning experiments with comprehensive metrics tracking.
Dataset Structure
The dataset contains the following columns:
experiment_id: Unique identifier for each experiment
name: Human-readable name for the experiment
description: Detailed description of the experiment
created_at: Timestamp when the experiment was created
status: Current… See the full description on the dataset page: https://huggingface.co/datasets/Tonic/trackio-experiments.shifaa_experimentsisoflop-experiments
IsoFLOP Scaling Law Experiments
Curated collection of IsoFLOP curve data from 6 experiments, standardized to a common schema.
This dataset is associated with the paper Problems with Chinchilla Approach 2: Systematic Biases in IsoFLOP Parabola Fits.
Project Page: https://openathena.ai/scaling-law-analysis
Data Extraction & Prep: Open-Athena/scaling-law-analysis
Scaling Law Estimation: Open-Athena/vpnls
Schema
Field
Type
Description
source
string
Data source… See the full description on the dataset page: https://huggingface.co/datasets/open-athena/isoflop-experiments.