datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
looped-qwen-v2-artifactsharbor-swesmith-rl-artifacts
Harbor SWE-Smith 强化学习数据产物
本数据集是 Harbor Qwen 工具调用代码智能体强化学习项目使用的冻结任务集，服务于 GRPO、原生价值模型/GAE PPO、训练过程诊断和统一协议评测。
项目已于 2026 年 8 月 30 日完成 P0 评测并进入阶段性归档。本数据集用于保留实验所依赖的数据切分、任务执行文件和审计信息，不代表新的通用代码能力基准。
数据概况
切分
任务数
训练集
187
验证集
42
测试集
38
合计
267
数据覆盖 89 个上游代码仓库。三个切分之间同时执行任务标识和仓库级隔离检查。
正式数据集名称：
swesmith-curated-grpo-267-v1
冻结切分的语义摘要：
ae5df9a3f4a3fc8af44fac420b36529e283839e1bd3de9daba65d5bcda51447d
该值来自 split-manifest.json 的 sha256 字段，用于标识切分语义，不等同于该文件本身的字节级… See the full description on the dataset page: https://huggingface.co/datasets/keryszhan/harbor-swesmith-rl-artifacts.tau2-uq-artifacts
tau2-bench UQ Artifacts
Interaction trajectories and token-level log-probability measurements from conversational customer service agent evaluations on tau2-bench, collected as part of the uncertainty quantification (UQ) pipeline. Used for analyses in the paper "Uncertainty Quantification in LLM Agents: Foundations, Emerging Challenges, and Opportunities" under the agentuq codebase.
Dataset Overview
This dataset contains two types of artifacts:
Trajectories --… See the full description on the dataset page: https://huggingface.co/datasets/changdae/tau2-uq-artifacts.public-agent-coordination-artifacts
Public Agent Coordination Artifacts
Real, complete edits and posts that AI agents left on public wikis and paste sites —
collected as open evidence for studying how autonomous agents use shared online spaces to
remember things, signal each other, and coordinate. It's the behavior spotlighted by the
mid-2026 OpenAI–Hugging Face agent incident,
here as raw public data researchers can actually inspect — plus a small, hand-reviewed map
of how specific artifacts relate.… See the full description on the dataset page: https://huggingface.co/datasets/leonidas1712/public-agent-coordination-artifacts.backup-drmas-noshare-8b-historical-run-artifacts
Historical DrMAS Noshare 8B run artifacts
Public backup of the historical drmas-checklist-bs16-n4-c1-agent Qwen3-8B run. This was a Noshare topology with separately trained Tool Caller and Tool Simulator agents (world_size=8).
Contents
rollout_dumps/: 470 training rollout JSONL files
val_dumps/: 47 validation JSONL files
Four training logs
latest_checkpointed_iteration.txt
522 business files and 5,189,359,695 logical bytes in total
Model scope
The… See the full description on the dataset page: https://huggingface.co/datasets/xuzishan/backup-drmas-noshare-8b-historical-run-artifacts.crooked-nebula-artifactspostdyn-artifactsvqa-cmsv-benchmark
VQA-CMSV Benchmark Data Package
This repository contains annotation splits for VQA v2-CMSV, GQA-CMSV, and VG-CMSV, plus patch-mask NPZ files used for mask supervision experiments.
Contents
data/vqa_v2_cmsv/train.json, data/vqa_v2_cmsv/val.json, data/vqa_v2_cmsv/test.json
data/gqa_cmsv/train.jsonl, data/gqa_cmsv/val.jsonl, data/gqa_cmsv/test.jsonl
data/vg_cmsv/train.jsonl, data/vg_cmsv/val.jsonl, data/vg_cmsv/test.jsonl
masks/vqa_v2_cmsv_masks.npz
masks/gqa_cmsv_masks.npz… See the full description on the dataset page: https://huggingface.co/datasets/as-benchmark-artifacts/vqa-cmsv-benchmark.grok-1-dissect-artifacts
Grok-1 structural parse artifacts (personal research)
Personal research tooling output — not a product release and not a hybrid-quantization
implementation. Structural parse of open Grok-1 checkpoint shards (tensor inventory, MoE
expert layout, routing-critical tensors, conversion policies, reports) produced by the
rmems/xai-dissect weight parser.
This is not a copy of the model weights, not a finetune, not an inference
runtime, and does not implement hybrid quantization.… See the full description on the dataset page: https://huggingface.co/datasets/rmems/grok-1-dissect-artifacts.backup-drmas-caller-only-8b-historical-run-artifacts
Historical DrMAS Caller-only 8B run artifacts
Public backup of the historical drmas-checklist-bs16-n4-c1-httpx-caller_only Qwen3-8B run. The Tool Caller was trained while the Tool Simulator used an external frozen Qwen3-8B service (world_size=8).
Contents
rollout_dumps/: 470 training rollout JSONL files
val_dumps/: 47 validation JSONL files
latest_checkpointed_iteration.txt
518 business files and 945,040,414 logical bytes in total
Model scope
The… See the full description on the dataset page: https://huggingface.co/datasets/xuzishan/backup-drmas-caller-only-8b-historical-run-artifacts.stillwarm-kv-cache-artifact
A downloadable KV-cache save file — with the honest math
One llama-server slot save: the first 8,192 Llama-tokens of Frankenstein
(public domain), prefilled by Qwen2.5-7B-Instruct Q4_K_M (Apache-2.0 model —
chosen over Llama specifically for artifact licensing) and saved with a
stillwarm sidecar.
This file is USELESS unless your setup matches the sidecar exactly:
field
value
llama.cpp build
b9871 (ef2d770117db45b05aa7ecd1b0acca36370c5470) — advisory: ±5 weeks measured… See the full description on the dataset page: https://huggingface.co/datasets/vimalnakrani/stillwarm-kv-cache-artifact.persona-artifactsvscode-issue-rag-artifactsnyc-restaurant-artifactsmicroduck-video-artifactsegret-artifactsicml-2026-30204-reproduction-artifacts
ICML 2026 #30204 reproduction artifacts
Artifacts for Networked Information Aggregation for Binary Classification (arXiv:2605.01082; OpenReview mrtg4NmvAe).
Downloads
icml-2026-30204-repro-bundle.tar.gz — complete 50-file reproduction bundle
MANIFEST.sha256 — checksums of files inside the unpacked bundle
poster.pdf — gate-verified 24x36 inch reproduction poster
summary.json — authoritative aggregate numerical results
Evidence_Memo.md — claim-by-claim source and… See the full description on the dataset page: https://huggingface.co/datasets/FloorIsAwake/icml-2026-30204-reproduction-artifacts.Latin-OCR-Artifacts
Latin sentences sourced from The Latin Library, converted to images, were subsequently degraded via OCRODEG.
OCR (via Kraken and Tesseract) transcriptions were generated. The dataset was then augmented with several synthetic noise patterns in order to emulate the more severe corruption found in many older digitizations.
If you use this in your work, please cite:
@misc{mccarthy2025LACOROCR,
author = {McCarthy, A. M.},
title = {{Latin OCR Artifacts}},
year = {2025}… See the full description on the dataset page: https://huggingface.co/datasets/aimgo/Latin-OCR-Artifacts.high-temp-refusal-probe-artifactsrepro-grace-artifacts
Artifacts — reproduction of GRACE (ICML 2026, OpenReview tSZaHvpxCd)
Raw outputs for the logbook at
https://huggingface.co/spaces/rakshi-the-neural-nexus/repro-gradient-based-causal-tree-ensembles-hte
File
What it is
run_grace.py
the driver that produced every cell: GRACE + four scikit-learn baselines (claim 1) and the nn.Linear -> GRACE_layer swap (claim 2)
results/grace.jsonl
one line per (method, dataset, seed) cell, 60 cells, as written during the run… See the full description on the dataset page: https://huggingface.co/datasets/rakshi-the-neural-nexus/repro-grace-artifacts.nyc-restaurant-artifactsfr-gentle-artifactecho-ml-artifactsflowos-eval-artifactsmultimodal-image-text-retrieval-artifactsmamba-mini-project-artifactsartifactscsvae