Corollary

Research

  • New investigation
  • Papers

Library

  • Catalog

Account

  • Jobs
  • Billing
  • Developers
SettingsResearch demo · not for clinical or commercial use
Corollary
  1. Catalog

Library

Find the model or dataset for the question you have.

Ask in plain language and we will work out where the answer lives — or filter the index yourself. Almost everything here runs on our servers, from this page, with no setup.

Models
300
Datasets
248
Organisations
131

Labs, institutes and companies

Ready to run
21

Including 6 that run free and instantly

Looking for a particular value, not a particular dataset?Search schemas, rows and column statistics inside the data itself.

Things that work

Filtering

48 entries in Scientific Reasoning

Showing 48 of 48

aimo-validation-aimeAI-MOAIME I/II problems reformatted for AIMO challenge validation — 15-question integer-answer format, covering competition math at difficulty levels 5–9.Competition MathExplore
aimo-validation-amcAI-MOAMC 10/12 competition problems reformatted for AIMO challenge validation, covering algebra, geometry, and number theory at difficulty levels 1–5.Competition MathExplore
aimo-validation-math-level-4AI-MOLevel-4 MATH benchmark problems (pre-calculus difficulty) used for AIMO challenge validation and fine-grained model evaluation.Math ProblemsExplore
aimo-validation-math-level-5AI-MOLevel-5 MATH benchmark problems (highest difficulty) used for AIMO challenge validation and measuring the ceiling of model mathematical reasoning.Math ProblemsExplore
AirfRANS_originalPLAID-datasetsOriginal AirfRANS airfoil RANS simulation dataset — graph-structured CFD over NACA airfoils for benchmarking physics-informed and graph neural networks.Aerodynamics CFDExplore
BioMysteryBench-fullAnthropicFull BioMysteryBench evaluation set — challenging biology problems used to probe expert-level scientific reasoning in frontier models.Biology BenchmarkExplore
BioMysteryBench-previewAnthropicPreview slice of BioMysteryBench — challenging, expert-curated biology problems for evaluating AI scientific reasoning capability.Biology BenchmarkExplore
bioreason-pro-sft-reasoning-datawanglabReasoning trace dataset used to supervised-fine-tune BioReason-Pro — multimodal biological problems with rationales over genomic variants and pathway data.Biological Reasoning CorpusExplore
BixBenchfuturehouseBenchmark with 205 reproducible research questions paired with data capsules for AI evaluation.Research BenchmarkExplore
ChemBenchjablonkagroupManually curated benchmark of 3,000+ chemistry and materials science questions across spectroscopy, reactivity, synthesis, and property prediction for evaluating LLMs.Chemistry BenchmarkExplore
chempile-captionjablonkagroupImage-to-text dataset of chemistry figures (molecular structures, reaction schemes, plots) with expert captions for training multimodal chemistry models.Chemistry CaptioningExplore
chempile-codejablonkagroupCurated chemistry-relevant code (RDKit, ASE, simulation tooling) drawn from The Stack — supports training models that can read and write computational chemistry workflows.Chemistry Code CorpusExplore
chempile-educationjablonkagroupEducational chemistry corpus — multiple-choice and open-ended items spanning introductory through graduate chemistry for assessing model educational capability.Chemistry Education CorpusExplore
chempile-instructionjablonkagroupInstruction-tuning corpus for chemistry — curated Q&A and dialogue traces drawn from chemical literature and educational sources for training chemistry-specialist LLMs.Chemistry Instruction CorpusExplore
chempile-liftjablonkagroupChemPile-LIFT — large-scale language-modelling dataset combining curated chemistry literature and structured chemical knowledge for foundation-model pretraining.Chemistry PretrainingExplore
chempile-reasoningjablonkagroupMulti-step chemistry reasoning corpus — open-domain QA, NLI, and multiple-choice items with chains of reasoning for training and evaluating chemical reasoning models.Chemistry Reasoning CorpusExplore
codeparrot_megatronADSKAILabMegatron-formatted CodeParrot release used for large-scale code language-model pretraining experiments at Autodesk AI Lab.Code PretrainingExplore
CombiBenchAI-MOCombinatorics problems drawn from AMC, AIME, and olympiad competitions, formalised for benchmarking discrete-mathematics reasoning in language models.CombinatoricsExplore
dataset_metallicglass_rc_llmfoundry-mlLLM-extracted critical cooling rate data for metallic glasses — text-mined complement to the structured Rc dataset.Metallic GlassExplore
equational-theories-benchmarkSAIRfoundationFull benchmark suite of equational theory problems spanning algebraic structures, designed to evaluate formal reasoning capabilities of AI models.Mathematical ReasoningExplore
equational-theories-selected-problemsSAIRfoundationCurated selection of equational theory problems for benchmarking LLM mathematical reasoning and automated theorem proving.Mathematical ReasoningExplore
ether0-benchmarkfuturehouseChemistry reasoning benchmark covering SMILES-based tasks including reaction prediction, retrosynthesis, and molecular property estimation for evaluating chemistry LLMs.Chemistry BenchmarkExplore
frontierscienceopenaiFrontier science evaluation benchmark probing model capabilities on expert-level reasoning across natural sciences — designed to surface what AI systems can and cannot do at the research frontier.Scientific Reasoning BenchmarkExplore
genomic-niahHuggingFaceBioGenomic Needle-in-a-Haystack — long-context evaluation for genomic LLMs, measuring retrieval and reasoning over long DNA sequences.Long-Context BenchmarkExplore
GeometryLeanBenchAI-MOGeometry theorem proving problems formalised in Lean 4, covering Euclidean, affine, and metric geometry for automated reasoning evaluation.Theorem ProvingExplore
healthbenchopenaiRealistic multi-turn health conversations graded against physician-written rubrics across multiple axes (accuracy, completeness, communication) — an open evaluation benchmark for AI assistants in medicine.Medical BenchmarkExplore
healthbench-professionalopenaiProfessional-graded subset of HealthBench: physician evaluators score model responses to clinically realistic conversations, targeting expert-level health assessment.Medical BenchmarkExplore
keggwanglabKEGG pathway entries paired with variant annotations for training and evaluating multimodal biological reasoning models (used by the BioReason work).Biological ReasoningExplore
Kimina-Prover-PromptsetAI-MOPrompt-set for training and evaluating Kimina, a Lean 4 theorem prover that uses reinforcement learning over formal mathematical proofs.Theorem ProvingExplore
lab-benchfuturehouseLanguage Agent Biology Benchmark - 8 categories of scientific research tasks including cloning, figures, and protocols.Research BenchmarkExplore
LLM-narrative-planning-tasksetADSKAILabNarrative planning task set for evaluating LLM planning and reasoning over multi-step design and engineering scenarios.Planning BenchmarkExplore
MedDialogOpenMedDoctor-patient medical dialogue dataset for training and evaluating clinical conversation models — covers triage, symptom checking, and diagnostic reasoning.Medical DialogueExplore
medical-o1-reasoning-SFTFreedomIntelligenceMedical chain-of-thought reasoning dataset (o1-style) for supervised fine-tuning of medical LLMs — one of the most-liked medical training corpora on Hugging Face (1000+ likes).Medical Reasoning CorpusExplore
medical-o1-verifiable-problemFreedomIntelligenceVerifiable medical reasoning problems with checker functions — supports RL/reward-model training for medical-LLM alignment beyond static SFT.Medical RL Reward DataExplore
Medical-Reasoning-SFT-MegaOpenMedLarge supervised fine-tuning corpus for clinical reasoning — multi-step medical question-answer chains with rationales for training instruction-following medical LLMs.SFT Reasoning CorpusExplore
minif2f_testAI-MOTest set for miniF2F formal mathematics benchmark.Theorem ProvingExplore
miriad-4.4Mmiriad4.4M-example medical reasoning subset of MIRIAD — earlier release used for benchmarking medical instruction-tuning workflows.Medical Reasoning CorpusExplore
miriad-5.8Mmiriad5.8M-example medical instruction-tuning and reasoning corpus curated from clinical literature for training healthcare LLMs at scale.Medical Reasoning CorpusExplore
NuminaMath-CoTAI-MONuminaMath with Chain-of-Thought reasoning annotations.Math ProblemsExplore
NuminaMath-TIRAI-MONuminaMath with Tool-Integrated Reasoning annotations.Math ProblemsExplore
olympiadsAI-MOOlympiad-level mathematical problems collected from international and national competitions, formatted for training and evaluating mathematical reasoning models.Math Reasoning CorpusExplore
olympiads-refAI-MOExtended reference set of olympiad problems with verified step-by-step solutions, used for Chain-of-Thought and formal reasoning training.Competition MathExplore
peS2oallenaiApproximately 40M cleaned, filtered, and formatted open-access academic papers derived from S2ORC — a large multi-domain pretraining corpus for science-aware language models, spanning biology, chemistry, engineering, computer science, and physics.Pretraining CorpusExplore
principia-benchfacebookCurated benchmark of challenging STEM problems requiring multi-step reasoning, quantitative analysis, and domain knowledge across natural sciences.STEM BenchmarkExplore
principia-collectionfacebookLarge-scale STEM reasoning dataset from Meta covering mathematics, physics, chemistry, and biology problems for training and evaluating scientific reasoning in language models.STEM ReasoningExplore
spiqagoogleScientific Paper Image Question Answering benchmark requiring multimodal reasoning over figures, charts, and diagrams from research papers across scientific domains.Scientific BenchmarkExplore
Zero-To-CAD-100kADSKAILabCurated 100K-example subset of Zero-To-CAD — useful for benchmarking and lightweight fine-tuning of CAD-from-image models.CAD Vision-Language CorpusExplore
Zero-To-CAD-1mADSKAILab1M paired image-and-CAD-program examples for training vision-language models that synthesise parametric CAD from images.CAD Vision-Language CorpusExplore