datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
atlas-22-prefix-accuracy-curves-neutral-vs-submit
ATLAS report 22: prefix accuracy curves k = 1..8 on the neutral selector and on the neutral-text submit surface
Complete raw products of ATLAS rl-training report 22 (GitHub issue #47).
Report 21's two ATLAS-free selector surfaces, run over every identity prefix
k = 1..8 of every question of the canonical LiveCodeBench (175) and GPQA (198)
validation sets: 373 questions x 8 prefixes x 2 surfaces = 5968 states, one
greedy deliberative rollout each, on the same Qwen3-14B checkpoint… See the full description on the dataset page: https://huggingface.co/datasets/t2ance/atlas-22-prefix-accuracy-curves-neutral-vs-submit.train_v6_filteredThis dataset is part of the paper Prefix Sliding for efficient test-time scaling. It contains training data for reinforcement learning to enable long-horizon reasoning.
Code is available at: https://github.com/Muennighoff/prefix-sliding
alfworld-expert-prefix-rollouts
ALFWorld Expert-Prefix Rollout Landscape
This dataset measures how a frozen language-model actor's probability of
solving an ALFWorld task changes after replaying different-length prefixes of a
successful expert trajectory.
The collection contains all 3,553 ALFWorld training tasks from the Agent-G2 SFT
data. Eight independent actor rollouts were sampled from the initial state for
every task. For the 2,307 low-signal tasks with at most one root success, eight
additional… See the full description on the dataset page: https://huggingface.co/datasets/YYYYYYibo/alfworld-expert-prefix-rollouts.task1319_country_by_barcode_prefix
Dataset Card for Natural Instructions (https://github.com/allenai/natural-instructions) Task: task1319_country_by_barcode_prefix
Additional Information
Citation Information
The following paper introduces the corpus in detail. If you use the corpus in published work, please cite it:
@misc{wang2022supernaturalinstructionsgeneralizationdeclarativeinstructions,
title={Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks}… See the full description on the dataset page: https://huggingface.co/datasets/Lots-of-LoRAs/task1319_country_by_barcode_prefix.prefixbench
PrefixBench JSONL Datasets
These datasets generate deterministic prompts for testing KV prefix caching behavior in LLM inference servers such as vLLM and SGLang. The prompts use controlled shared prefixes plus small unique suffixes so benchmark clients can compare cache reuse, latency, and throughput across server configurations.
Files
shared_schema_1k.jsonl: Simple shared-prefix benchmark. Every request reuses the same extraction instruction, JSON schema, and few-shot… See the full description on the dataset page: https://huggingface.co/datasets/jaytonde05/prefixbench.finops_token_cost_prefix_cache_terminator_teaser
🚀 Enterprise FinOps - AI Token Cost & Prefix-Cache Terminator (Evaluation Teaser)
⚡ Official Free Evaluation Teaser (50 Verified Multi-Turn Scenarios)🏆 Get the Full Production Package (500 Samples) & Commercial EULA on Gumroad:👉 Enterprise FinOps - AI Token Cost & Prefix-Cache Terminator on Gumroad🏷️ Use coupon code LAUNCH20 for 20 € off at checkout!
📦 What is Inside the Full Production Package:
500 Verified FAANG v2.0 Scenarios (100% AST-Valid Python)… See the full description on the dataset page: https://huggingface.co/datasets/emgena/finops_token_cost_prefix_cache_terminator_teaser.clean-gsm8k-aug-prefix
Clean GSM8K-Aug-Prefix
Overview
The Clean GSM8K-Aug family is a revised version of
whynlp/gsm8k-aug and
whynlp/gsm8k-aug-nl.
It retains the original question, steps, and answer schema while removing
or repairing examples with incomplete or inconsistent calculation traces.
Four representations of the same aligned questions and answers are available:
Dataset
Step representation
cs-giung/clean-gsm8k-aug
Infix arithmetic expressions… See the full description on the dataset page: https://huggingface.co/datasets/cs-giung/clean-gsm8k-aug-prefix.iris-prefix-cache-benchmark
vLLM Iris Prefix Caching Benchmark Dataset
This dataset is specifically designed to test and benchmark the Automatic Prefix Caching feature in vLLM, using the technical announcement of vLLM Semantic Router v0.1 (Iris) as the shared context.
Dataset Structure
The dataset contains 20 prompts. Each prompt consists of:
Shared Prefix: The technical overview of the vLLM Semantic Router Iris release (~500 tokens).
Unique Suffix: A specific technical question based on the text.… See the full description on the dataset page: https://huggingface.co/datasets/jaytonde05/iris-prefix-cache-benchmark.musical-instruments-sft-selected-prefix-20260923
Musical Instruments selected-checkpoint SFT prefix
The exact 6,720 training examples consumed by full-parameter Qwen3-4B SFT checkpoint 420. Rows retain original candidate IDs and zero-based shuffled epoch positions. The complete one-epoch run used 8,937 examples, but the selected checkpoint consumed only this prefix. Reconstructed from saved input order, seed 42, explicit Python shuffle,420 updates × microbatch 2 × accumulation 8; no DataLoader/prefetch or resume.
Source:… See the full description on the dataset page: https://huggingface.co/datasets/flavianv/musical-instruments-sft-selected-prefix-20260923.rlve_offline_20K_POPE_prefix_pass1_qwen3-1.7b
RLVE offline-20K POPE-prefix completions — Qwen3-1.7B (pass1)
Prefix-conditioned completions generated by Qwen3-1.7B over the
rlve_offline_20K_POPE_prefix prompt set (20000 records, 1 sample/prompt).
Produced by SLURM job 6580578 (vLLM, tp=2), 2026-06-15.
Fields
index, sample_id, prompt, prefix, response, answer, rewards
⚠️ Caveat on rewards
The inline rewards field is all 0.0 — this is the known inline-Gym-verifier
artifact (same as the old… See the full description on the dataset page: https://huggingface.co/datasets/CL-From-Nothing/rlve_offline_20K_POPE_prefix_pass1_qwen3-1.7b.ptdbench-reward-design-reward-prefix-product-mod-distinct-permutation-011-dataset
PTDBench dataset snapshot: reward_prefix_product_mod_distinct_permutation_011
This repository stores the immutable runtime dataset snapshot for one
materialized PTDBench task. It intentionally excludes model weights and
training checkpoints.
PTDBench family: reward_design
Source evaluation metric: eval/HELD-OUT_ENVIRONMENTS_128
Provenance: RLVE repository snapshot under its MIT license; bundled upstream benchmark notices remain applicable.
License: MIT
The artifact manifest… See the full description on the dataset page: https://huggingface.co/datasets/LIF1014/ptdbench-reward-design-reward-prefix-product-mod-distinct-permutation-011-dataset.ptdbench-reward-design-reward-prefix-sum-mod-distinct-permutation-010-dataset
PTDBench dataset snapshot: reward_prefix_sum_mod_distinct_permutation_010
This repository stores the immutable runtime dataset snapshot for one
materialized PTDBench task. It intentionally excludes model weights and
training checkpoints.
PTDBench family: reward_design
Source evaluation metric: eval/HELD-OUT_ENVIRONMENTS_128
Provenance: RLVE repository snapshot under its MIT license; bundled upstream benchmark notices remain applicable.
License: MIT
The artifact manifest… See the full description on the dataset page: https://huggingface.co/datasets/LIF1014/ptdbench-reward-design-reward-prefix-sum-mod-distinct-permutation-010-dataset.