datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
openfoam
OSWorld-Science · OpenFOAM 全量归档 / full project archive
这个仓库是 OSWorld-Science 基准 OpenFOAM 部分的完整归档：题目、真值与阈值、评分器、虚拟机（guest 桌面镜像）的构建材料与镜像本身、
每一次 agent 运行的轨迹、审核者的复现工具与算例、以及产生这一切的全部 Claude Code 会话记录。
仓库是 gated（人工审批）：任何人能看到文件名，只有被批准的账号能读内容。
Access policy / 访问约定：内容含答案与阈值，属未发布基准的答案侧材料，只对协作者开放；请勿再分发、勿用于训练。
Layout / 目录
path
what
来源与说明
osworld-openfoam/tasks/
题面、网格、交付规格（每题一目录）
由本机 6 小时定时同步维护
osworld-openfoam/harness/, scripts/, env/, docs/
评分器（四道门… See the full description on the dataset page: https://huggingface.co/datasets/daipath/openfoam.leanflow-dualscale-openfoam-nse-benchmark
LeanFlow DualScale vs. OpenFOAM (icoFoam) Navier-Stokes Benchmark & Regularity Audit
Cross-Solver Verification: Mathematics, Numerics, and Physics
Authors: Xavier Callens / RunuX AI Runtime & SocrateAIGitHub Repository: xaviercallens/runux-ai-runtimeRelease Tag: v0.5.0-dualscale-nseFormal Verification: Lean 4 Formal Specification (spec/RunuxSpec/LeanFlow/)
1. Overview & Scope
This dataset and research bundle provides an exhaustive… See the full description on the dataset page: https://huggingface.co/datasets/callensxavier/leanflow-dualscale-openfoam-nse-benchmark.openfoam-Agent-Dataset
foam-openfoam-agent — Validated OpenFOAM Case Dataset
A dataset of natural-language CFD prompts paired with validated, converged
OpenFOAM v2412 case files. Every case in this dataset was generated by the
foamllm3 agent pipeline and verified by actually
running the solver to convergence; only converged runs with quality
score ≥ 0.5 are included.
The dataset ships in two formats from the same source:
Config
File
Granularity
Records
Use
per_file
foam_openfoam_dataset.jsonl… See the full description on the dataset page: https://huggingface.co/datasets/arungovindneelan/openfoam-Agent-Dataset.openfoam-casegen-dataOpenFOAM-bpOpenfoam-Osworld_Benchmark
OpenFOAM OSWorld Benchmark
This gated dataset contains OSWorld-style, long-horizon OpenFOAM and ParaView
tasks. Each release task combines deterministic CFD evidence and engineering
decision criteria with a substantive native ParaView state and scientific image
deliverable. CLI assistance is allowed, but the final GUI evidence is graded.
Current release tasks
The current five-task set is:
of_battery_coldplate_release_review_v4… See the full description on the dataset page: https://huggingface.co/datasets/daipath/Openfoam-Osworld_Benchmark.openfoam_tanzim