| { |
| "@context": { |
| "@language": "en", |
| "@vocab": "https://schema.org/", |
| "sc": "https://schema.org/", |
| "cr": "http://mlcommons.org/croissant/", |
| "dct": "http://purl.org/dc/terms/", |
| "conformsTo": "dct:conformsTo", |
| "containedIn": "cr:containedIn", |
| "fileObject": "cr:fileObject", |
| "fileSet": "cr:fileSet", |
| "field": "cr:field", |
| "recordSet": "cr:recordSet", |
| "dataType": "cr:dataType", |
| "examples": "cr:examples", |
| "extract": "cr:extract", |
| "source": "cr:source", |
| "annotation": "cr:annotation", |
| "references": "cr:references", |
| "format": "cr:format", |
| "jsonPath": "cr:jsonPath", |
| "regex": "cr:regex", |
| "isLiveDataset": "cr:isLiveDataset", |
| "rai": "http://mlcommons.org/croissant/RAI/", |
| "prov": "http://www.w3.org/ns/prov#" |
| }, |
| "@type": "sc:Dataset", |
| "name": "ORC-bench", |
| "description": "ORC-bench is a benchmark dataset for reasoning over power systems, optimal power flow, control decisions, and cybersecurity tasks. The dataset contains 11 JSON test files spanning 7 tasks, including topological path finding, topological connectivity, linear power flow, contingency analysis, agentic grid control, optimal power flow optimization, credit scoring, botnet detection, and phishing URL reasoning.", |
| "url": "https://huggingface.co/datasets/orc-bench", |
| "sameAs": "https://huggingface.co/datasets/orc-bench", |
| "license": "https://creativecommons.org/licenses/by/4.0/", |
| "dct:conformsTo": "http://mlcommons.org/croissant/1.1", |
| "citeAs": "ORC-bench dataset (2026). https://huggingface.co/datasets/orc-bench", |
| "creator": { |
| "@type": "sc:Organization", |
| "name": "ORC-bench" |
| }, |
| "publisher": { |
| "@type": "sc:Organization", |
| "name": "ORC-bench" |
| }, |
| "datePublished": "2026-05-06", |
| "version": "1.0.0", |
| "sdVersion": "1.0", |
| "keywords": [ |
| "power-systems", |
| "electricity", |
| "graph-reasoning", |
| "optimization", |
| "optimal-power-flow", |
| "benchmark", |
| "json", |
| "tabular", |
| "cybersecurity", |
| "credit-scoring", |
| "urls", |
| "ieee", |
| "pglib-opf", |
| "pypsa", |
| "networkx", |
| "pytorch-geometric" |
| ], |
| "distribution": [ |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_1/topological_path_test.json", |
| "name": "task_1/topological_path_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_1/topological_path_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "17487470 B", |
| "sha256": "b07fef242644cfea8afdeaf8e9adecd71331dc0b85b625494f0e0a87c609090d" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_2/topological_connectivity_test.json", |
| "name": "task_2/topological_connectivity_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_2/topological_connectivity_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "16840750 B", |
| "sha256": "0484e46708b2b89cf1368d6e7f96e7d8f799511144453f4facf652761f688c01" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_3/physical_test.json", |
| "name": "task_3/physical_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_3/physical_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "16655185 B", |
| "sha256": "65ff3da0ca390100fa6ac24829ea1308bcb1a0994313b0a6788b839d88601ecd" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_4/counterfactual_test.json", |
| "name": "task_4/counterfactual_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_4/counterfactual_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "18578608 B", |
| "sha256": "7cf0e58128477085586314441ca4f120ba3e92ca5a4c0e461e71673200fa47a7" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_5/agentic_test.json", |
| "name": "task_5/agentic_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_5/agentic_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "5127829 B", |
| "sha256": "cc289e4ef1b28878df9f2a394384d09536f5b1600ceb41b18494115c512c3024" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_6/opf_pglib_opf_case14_ieee_test.json", |
| "name": "task_6/opf_pglib_opf_case14_ieee_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_6/opf_pglib_opf_case14_ieee_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "142810229 B", |
| "sha256": "dbee301de6ba1420553c56bcfc703fdb87d3fde48cb01cab407b0bf353dbe3db" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_6/opf_pglib_opf_case30_ieee_test.json", |
| "name": "task_6/opf_pglib_opf_case30_ieee_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_6/opf_pglib_opf_case30_ieee_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "133940685 B", |
| "sha256": "c3171cdd830b6efae09552d446788bb834df475c53c388f2e69ef06f75887ea2" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_6/opf_pglib_opf_case118_ieee_test.json", |
| "name": "task_6/opf_pglib_opf_case118_ieee_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_6/opf_pglib_opf_case118_ieee_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "175450487 B", |
| "sha256": "1347d1f4e371b31cdcbe9483331ec8d1842128e6c815c91d8ca53ac0c1fba424" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_7/ctu_test.json", |
| "name": "task_7/ctu_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_7/ctu_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "407728077 B", |
| "sha256": "d35b186c3bb5874bd1ea7dadfc259c96cca5e36ac5fe37219a3ddee95bf6a62a" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_7/lcld_test.json", |
| "name": "task_7/lcld_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_7/lcld_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "122011294 B", |
| "sha256": "bdecac54b51b5ac3e2b4971580bdd7b86a3ef3c2a143788b87608ec8e100bacc" |
| }, |
| { |
| "@type": "cr:FileObject", |
| "@id": "task_7/url_test.json", |
| "name": "task_7/url_test.json", |
| "contentUrl": "https://huggingface.co/datasets/orc-bench/resolve/main/task_7/url_test.json", |
| "encodingFormat": "application/json", |
| "contentSize": "1846610 B", |
| "sha256": "4897f6e332717a74c927e4cec44d8d2d2a5f37436a6c0237aae65d9574a3778a" |
| } |
| ], |
| "rai:dataLimitations": "ORC-bench focuses on static, simplified settings (e.g., small IEEE/PGLib grids and linearized/steady-state power-flow) and curated cybersecurity/credit/URL corpora, so results may not transfer to larger, dynamic, noisy, or real-time environments. It is not recommended for production grid operations, safety-critical control, security incident response, or real-world lending/eligibility decisions without independent validation and regulatory review.", |
| "rai:dataBiases": "The grid tasks center on canonical IEEE/PGLib test cases that represent well-studied topologies and operating conditions, while the URL and botnet corpora are susceptible to time-based drift and family/host skew, and credit data (e.g., Lending Club) can embed selection and survivorship biases. These biases may cause models to overfit to specific network motifs or attacker behaviors.", |
| "rai:personalSensitiveInformation": "No direct Personal & Sensitive Information like names or SSNs is expected in these test JSONs.", |
| "rai:dataUseCases": "The dataset measures multi-step reasoning over power-system topology/physics and optimization (OPF), agentic grid control decision-making, and classification tasks in cybersecurity (botnets, phishing) and credit risk. It is validated as a benchmarking suite for task accuracy/feasibility on IEEE/PGLib OPF cases and widely used security/credit corpora.", |
| "rai:dataSocialImpact": "ORC-Bench aims to expose, rather than enhance, LLM capabilities in constrained reasoning. We anticipate broader societal impacts by helping practitioners avoid premature deployment of LLMs in high-stakes domains such as power grid operations, financial underwriting, and cyber-security where unsatisfied physical or domain constraints could translate into blackouts, discriminatory credit decisions, or missed intrusions. A potential negative impact is that improvements against the benchmark could lend over-confidence to operators who deploy LLMs autonomously in safety-critical infrastructure.", |
| "rai:hasSyntheticData": false, |
| "prov:wasDerivedFrom": [ |
| { |
| "@id": "https://console.cloud.google.com/storage/browser/gridopt-dataset", |
| "prov:label": "gridopt-dataset", |
| "sc:license": "CC BY 4.0", |
| "prov:wasAttributedTo": { |
| "@id": "https://arxiv.org/abs/2406.07234" |
| } |
| }, |
| { |
| "@id": "https://huggingface.co/datasets/serval-uni-lu/tabularbench", |
| "prov:label": "TabularBench", |
| "sc:license": "CC BY 4.0", |
| "prov:wasAttributedTo": { |
| "@id": "https://arxiv.org/abs/2408.07579" |
| } |
| } |
| ], |
| "prov:wasGeneratedBy": [ |
| { |
| "@type": "prov:Activity", |
| "prov:type": { |
| "@id": "https://www.wikidata.org/wiki/Q5227332" |
| }, |
| "prov:label": "Samples generation", |
| "sc:description": "Generating flows/contingencies and OPF solutions via open-source toolchains (e.g., PyPSA, NetworkX) and solver backends (tasks 1 to 5).", |
| "prov:wasAttributedTo": [ |
| { |
| "@type": "prov:Agent", |
| "@id": "salah_ghamizi", |
| "prov:label": "Salah Ghamizi", |
| "sc:description": "Research Scientist @ SnT, University of Luxembourg" |
| }, |
| { |
| "@type": "prov:Agent", |
| "@id": "fabien_bernier", |
| "prov:label": "Fabien Bernier", |
| "sc:description": "Postdoctoral Researcher @ SnT, University of Luxembourg" |
| } |
| ] |
| }, |
| { |
| "@type": "prov:Activity", |
| "prov:type": { |
| "@id": "https://www.wikidata.org/wiki/Q5227332" |
| }, |
| "prov:label": "Prompt formulation", |
| "sc:description": "Converting canonical grid cases to JSON prompts/labels", |
| "prov:wasAttributedTo": [ |
| { |
| "@type": "prov:Agent", |
| "@id": "fabien_bernier", |
| "prov:label": "Fabien Bernier", |
| "sc:description": "Postdoctoral Researcher @ SnT, University of Luxembourg" |
| }, |
| { |
| "@type": "prov:Agent", |
| "@id": "salah_ghamizi", |
| "prov:label": "Salah Ghamizi", |
| "sc:description": "Research Scientist @ SnT, University of Luxembourg" |
| } |
| ] |
| } |
| ] |
| } |
