Full Papers
Submissions must use the AI4Physics @ ICML 2026 LaTeX template, with a main body of up to eight pages, followed by unlimited pages for references and appendices, submitted as a single PDF.
Machine Learning/CORE Unranked
AI4Physics 2026
ICML 2026 Workshop on AI for Physics
Apr 01 2026Official workshop site Site reachable
AI4Physics 2026 is an ICML workshop bringing together researchers from machine learning and physics to advance AI-driven scientific research. It focuses on real-world applications in computational, theoretical, and experimental physics, including high-energy physics, quantum science, and cosmology, with an emphasis on physically grounded AI systems, reliable inference, and community-driven benchmarks.
Key deadlines
Verified September 9, 2026Full paper
May 8, 2026
AoE
Workshop timeline
Submission and decisions
Full paperKey deadline
May 8, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Physics-centric Scientific Reasoning with LLMs and Agents
Generating physically consistent hypothesesDeriving predictions with correct assumptions, units, and constraintsInterpreting outcomes from simulations or experimentsTool-augmented agentsRetrieval over scientific papers and codeStructured memory for multi-step derivations and verification
02
High-fidelity Generative and Surrogate Simulators for Physics
Learning-based simulators and emulators for complex physical processesPDE-governed dynamicsTurbulence, plasma systems, cosmology, detector-level simulationNeural operatorsPhysics-constrained generative modelsDifferentiable simulatorsHybrid solversError-controlled emulation strategiesLong-horizon rolloutsStiff dynamicsMultiscale couplingRare events
03
Inverse Problems and Systematic Inference
Recovering physical parameters, fields, or latent states from indirect measurementsLikelihood-free and simulation-based inferenceDifferentiable and amortized inferenceInverse designHandling nuisance parametersCalibration errorsSelection effectsSimulation-measurement mismatch
04
World Models, Extrapolation, and Transfer Across Regimes
Learned, stateful models of physical systemsMapping multimodal observations to compact latent representationsPredicting dynamics under partial observabilityExtrapolation to unseen regimesTransfer from simulation to experimentHybrid approaches combining physical structure with data-driven representationsEvaluation centered on out-of-distribution reliability
05
Experimental Data Scarcity, Bias, and Dataset-building for Physics
Lack of large, standardized, openly accessible physics resourcesDataset creationBenchmark designReproducible evaluationSynthetic data with controllable realismWeak supervisionTargeted data acquisitionActive learningAdaptive measurementAutonomous experimentation
Policies worth checking twice
- Double-blind review process
- Papers generated by AI or autonomous research systems will be desk-rejected
- First author of each submission must serve as a reviewer
- Submissions under review at other venues must comply with those venues' dual-submission and anonymity policies
- Accepted papers will not appear in formal archival proceedings (non-archival)
- All submissions must comply with the ICML Code of Conduct
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026