PaperlayerView exact revision on arXiv

Source-linked AI summary

Framework for an AI-Ready ARPES Community Database

Zhi-Xun Shen, Jin Gene Wong, Karen Lei, Leon Garcia, Felix Baumberger, Sergey Borisenko, Aaron Bostwick, Ting Cao, Tanny Chavez, Yulin Chen, Monika Choudhary, Andrea Damascelli, Dan Dessau, Thomas P. Devereaux, Sandy A. Ekahana, Alexei Fedorov, Atsushi Fujimori, Yucheng Guo, M. Zahid Hasan, Makoto Hashimoto, Yu He, Philip Hofmann, Zahid Hussain, Harold Y. Hwang, Zhurun Ji, Chunjing Jia, Timur K. Kim, Phil D. C. King, Takeshi Kondo, Alessandra Lanzara, Chun Lin, Donghui Lu, Matteo Michiardi, Sung-Kwan Mo, Brian Moritz, Laurenz Rettig, Eli Rotenberg, Takafumi Sato, Kyle M. Shen, Jonathan A. Sobota, Inna Vishik, Shuolong Yang, Ming Yi, Timothy Zaklama

arXiv:2610.15236v1cond-mat.mtrl-scicond-mat.str-elcond-mat.supr-concs.AI

TL;DR

ARPES data are difficult to interpret without experimental context, and publication-centered practices leave much raw data, code, and informal knowledge siloed or lost. The paper proposes a shared ecosystem and practical deposition steps; its proposal treats publication as a waypoint and identifies unresolved challenges in building the data foundation for broader AI use.

Abstract

from arXiv · show

Condensed matter stands to gain profoundly from widespread adoption of artificial intelligence (AI). However, its current means of data organization impedes this endeavor. As a premier measure of electronic structure underlying material properties, Angle-Resolved Photoemission Spectroscopy (ARPES) has accumulated over four decades of spectra, yet the community's publication-centered workflow preserves only a highly selected fraction of its experimental record. A shared data ecosystem could transform these fragmented measurements and their contexts into a persistent, collective scientific memory, creating immediate value for reproducibility and reuse while laying the foundation for AI-enabled discovery. In this Perspective, we report outcomes of a recent community workshop and propose a practical path toward such an ecosystem. Its benefits unfold along three timescales. In the near term, standardized, searchable, and citable datasets strengthen analysis pipelines independently of AI. In the intermediate future, community-validated tools operating across laboratories could enable broader reuse and increasingly closed-loop automated experiments. In the long term, a diverse corpus of measured electronic spectra could support foundation models for quantum matter. Looking forward, a diverse corpus of measured electronic spectra could support foundation models for quantum matter and new discoveries. We discuss lessons from existing scientific databases, the current state of AI in ARPES, as well as areas of community convergence on data organization, access, credit, governance, and a staged pilot program.

I. WHY ELECTRONIC STRUCTURE MATTERS, AND THE ARPES OPPORTUNITY

Electronic structure provides a microscopic basis for material properties, and ARPES probes it through momentum- and energy-resolved photoemission. But measured spectra depend on experimental context, while the publication-centered workflow loses data and knowledge; the proposed ecosystem would make publication a waypoint within a shared, persistent framework.

I. WHY ELECTRONIC STRUCTURE MATTERS, AND THE ARPES OPPORTUNITY

  • Electronic structure links quasiparticle behavior to macroscopic properties, and ARPES accesses it through energy- and momentum-resolved photoemission spectra.Conventional equilibrium ARPES primarily probes the occupied one-electron spectral function, modulated by matrix elements and experimental resolution.Source: §I · p. 3

A. ARPES data are extremely rich, but they come with inherent complexity

  • ARPES intensity is not the theoretical spectral function: experimental geometry, resolution, processing, and sample-surface conditions can make nominally identical spectra look substantially different.Measurements sample slices of high-dimensional momentum space whose locations are revealed in post-processing; preserving context is central to interpretation.Source: A. ARPES data are extremely rich, but… · p. 3

B. The current ARPES workflow is suboptimal

  • The linear workflow loses information at each step, and published figures select for successful, interpretable measurements while rarely preserving unsuccessful, ambiguous, or seemingly uninteresting spectra.Raw data and analysis code often remain localized to laboratories, and informal experimental knowledge can disappear when researchers leave the field.Source: B. The current ARPES workflow is… · pp. 3, 5
  • A community-wide ecosystem would place publication within a shared, persistent framework that includes unsuccessful and ambiguous measurements, before incompatible local conventions become entrenched.The proposal treats a sufficiently rich database as a way to consolidate siloed experimental data and context.Source: B. The current ARPES workflow is… · p. 5

C. Immediate opportunities of the ecosystem — before foundation models

Before foundation models, a shared ARPES ecosystem could preserve and make data findable and citable, while broadening reuse across measurements and material families. It could also help researchers revisit older data and identify spectra that merit further attention.

C. Immediate opportunities of the ecosystem — before foundation models

  • Standardized datasets would become durable, searchable, citable research objects, improving preservation of data held on aging storage or in obsolete formats.These immediate benefits do not depend on AI and could provide a durable basis for broader applications.Source: C. Immediate opportunities of the… · p. 5
  • A shared archive could make ineffective and ambiguous measurements useful as negative examples, support retrospective searches for phenomena recognized later, and surface anomalous spectra.The paper calls retrospective discovery of previously unrecognized phenomena in old experiments “precovery.”Source: C. Immediate opportunities of the… · p. 6
  • The shared-archive workflow closes the experimental loop with standardized ingest, validated tools, DOI-based sharing, and reuse by people and models.The diagram presents the change as organizational—a searchable index with provenance—not new hardware.Source: Figure 3 · p. 7
  • Making accumulated experience accessible across laboratories could help researchers compare measurements across chemically and structurally unrelated systems, beyond predefined material categories.The ecosystem could expose younger researchers to a broader range of measurements and materials than any one laboratory’s datasets.Source: C. Immediate opportunities of the… · pp. 6–7

III. FROM TODAY’S AI TOOLS TO MODERN AI-ENABLED SCIENCE

Existing databases show the value of durable shared records, but ARPES lacks the compact, widely agreed representation common in structural databases. The paper argues that an ARPES ecosystem will need standardized formats and benchmarks while retaining original data.

A. What other data ecosystems teach us

  • Condensed-matter precedents include the Materials Data Repository, XAS Data Library, and nmrXiv [39], alongside structural databases such as ICSD and PubChem.Source: A. What other data ecosystems teach us · p. 7
  • The Protein Data Bank grew into a durable record of more than a quarter-million experimentally determined structures; AlphaFold later used PDB structures, although the PDB was not created for that purpose.The paper presents the PDB as a precedent for a shared scientific record whose downstream uses can emerge decades later.Source: A. What other data ecosystems teach us · p. 7
  • Unlike structural databases, ARPES lacks a universal compact representation because intensity is entangled with experimental context and analysis depends on bespoke pipelines.The paper calls for standardized spectral formats and benchmarks, quality tiering, and retention of original data rather than only processed results.Source: A. What other data ecosystems teach us · pp. 7–8

B. Status of AI in ARPES today

Current ARPES AI spans analysis, representation, and experiment control, but generalization remains limited by narrow training data. The paper describes foundation-model approaches and their possible new capabilities, while emphasizing unresolved data and modeling challenges.

B. Status of AI in ARPES today

  • Current ARPES AI covers analysis, representation and generalization, and experiment control, including denoising, feature extraction, few-shot classification, and autonomous microARPES.The paper says these existing models could benefit from larger and more diverse ecosystem training data even before foundation models are built.Source: B. Status of AI in ARPES today · p. 8

C. Towards future foundation models

  • Models trained on narrow experimental or synthetic datasets struggle with unfamiliar band structures, detector conditions, or sample states; dependable out-of-distribution robustness requires broader, diverse data.The paper describes experimental data as essential for realism, validation, and reliable generalization.Source: C. Towards future foundation models · p. 8
  • A foundation-model approach pre-trains on large, diverse datasets for general representations, then adapts to specific tasks with smaller, high-quality experimental datasets.Pre-training may combine large synthetic corpora with broader experimental spectra, which capture complexities omitted by simulation.Source: C. Towards future foundation models · p. 8
  • A sufficiently broad corpus could support recognition of later-discovered phenomena, new classes of quantum matter, anomaly ranking, and multimodal links across experimental techniques and theory.The paper also identifies longer-term inverse questions connecting desired electronic structures to candidate materials.Source: C. Towards future foundation models · p. 8
  • Synthetic spectra offer scale but propagate simulator assumptions, while measured spectra are realistic but noisy and limited; the optimal architecture and synthetic-data balance remain unresolved.The paper identifies community involvement and a rich, well-organized data foundation—not a particular architecture—as central to realizing the vision.Source: C. Towards future foundation models · p. 8

IV. DESIGNING AN ARPES DATA ECOSYSTEM FOR REUSE AND DISCOVERY

The proposed ecosystem combines reusable data infrastructure with a staged approach to metadata, access, credit, and evaluation. A compact metadata core and flexible extensions support interpretation while preserving local acquisition practices and experimental context.

A. Organizing data and metadata

  • NeXus/HDF5 and its ARPES application NXmpes offer existing technical foundations aligned with FAIR principles, rather than requiring a new container from scratch.The authors also call for careful infrastructure redesign that considers these and other existing technical foundations and data-management principles.Source: A. Organizing data and metadata · p. 9
  • Use a compact mandatory metadata core to reconstruct experimental geometry and momentum coordinates, alongside extensible free-text context about samples, conditions, processing, and intent.Automated capture should record metadata by default, with hosting capacity planned for growing data volumes.Source: A. Organizing data and metadata · p. 9

B. Areas of convergence

  • Preserve raw measured coordinates and provenance, convert data on ingest, and begin with published datasets to lower the barrier to a shared archive.The mandatory metadata core should remain practical and small, while free-text annotations retain experimental judgment and decision context.Source: B. Areas of convergence · p. 9
  • Datasets should have persistent identifiers and explicit credit, support access tiers, and be hosted through pilot sites before permanent infrastructure is selected.Multiple independent pilot sites can estimate storage, conversion effort, and metadata completeness while surfacing practices for eventual integration.Source: B. Areas of convergence · pp. 9–10
  • Define benchmark tasks and evaluation criteria early so analysis and AI approaches can be compared systematically before claims about general-purpose foundation models.The steering committee is tasked with coordinating follow-up work and engaging the community on unresolved technical, policy, and governance questions.Source: §IV · p. 9, B. Areas of convergence · p. 10

C. Scaling up the program

The proposed program proceeds through pilots, evaluation, and then expansion, with each phase establishing outputs and goals before the next begins. It moves from published multi-laboratory data toward validated tools, broader participation, and possible foundation-model training.

C. Scaling up the program

  • Small-scale working demonstrations are proposed to resolve technical and community-buy-in differences before the program advances through three phases.Initial pilots are being set up by willing participants and begin with published datasets spanning laboratories, facilities, instruments, and local conventions.Source: C. Scaling up the program · p. 10
  • Phase 1 pilots should preserve native data, establish reusable converters, define minimal metadata for geometry and provenance, and assign persistent identifiers without imposing one experimental convention.The initial pilots are intended to surface early implementation problems.Source: C. Scaling up the program · pp. 10, 12
  • Phase 2 should test sustainability, fairness, and usefulness while estimating storage, securing hosting, agreeing on attribution and access policies, and measuring contributor burden.An end-to-end demonstration should connect automatic metadata capture during an experiment to automatic database upload.Source: C. Scaling up the program · p. 12
  • Phase 2 usefulness should be tested through outside-group reuse and benchmark tasks using held-out experimental test sets and conventional analysis as a baseline.Proposed tasks include cross-instrument dispersion extraction, uncertainty-aware denoising, and cross-material similarity search.Source: C. Scaling up the program · p. 12
  • Phase 3 would extend metadata capture to more facilities, add specialty measurement modes and private or prepublication staging, and broaden analysis and experiment-support tools.As corpus diversity and quality grow, benchmark-validated AI tools could feed back into acquisition and adaptive measurement, and ultimately support future ARPES foundation models.Source: C. Scaling up the program · p. 12

V. A PRACTICAL PATH FORWARD — AND LESSONS BEYOND ARPES

The authors argue that a community data ecosystem has scientific value beyond AI because experimental data and tacit context are vulnerable to loss. Preserving reusable measurements and their context could make the scientific record searchable and open measurements to questions not yet conceived.

V. A PRACTICAL PATH FORWARD — AND LESSONS BEYOND ARPES

  • Experimental condensed matter data exceed what papers convey, while interpretive knowledge remains tacit and vulnerable as researchers leave, instruments are decommissioned, and laboratories close.The authors argue communities should preserve measurements while they can still determine what context matters and how to record it.Source: §V · p. 12
  • Preserving measurements with interpretive context would retain ambiguities, failures, alternative interpretations, and conditions often omitted from publication, enabling later search and comparison across experiments and materials.The authors stress that future discoveries cannot be known in advance.Source: §V · p. 12

Supplementary Information for “Framework for an AI-Ready ARPES Community

  • The Supplementary Information documents the recommendations’ basis and implementation, including workshop methods, ARPES data context, metadata, AI applications, governance, and pilot deliverables.It also records unresolved decisions and individual discussion contributions without attribution.Source: Supplementary Information for… · p. 16

A. Format, participants, and sources

The Supplementary Information documents workshop inputs, how those inputs are interpreted, bibliometric methods, and reported group practices. It also records limitations in the source material and earlier curation efforts.

A. Format, participants, and sources

  • The 2026 Stanford conference included 23 invited talks and four structured discussion sessions; speakers addressed scientific goals and what would motivate data deposition.Workshop materials included slides, recap materials, the discussion-board export, and the main text.Source: A. Format, participants, and sources · p. 16

B. How the Day 2 board is interpreted

  • The Day 2 board export records 187 cards, 127 comments, 617 votes, 50 participants, and 48 voters, with anonymous cards.Votes could be multiple per person and merged-card totals aggregate component votes; counts indicate salience among a self-selected group, not broader-community consensus.Source: B. How the Day 2 board is interpreted · p. 16
  • The Supplement classifies recommendations as workshop-supported principles, proposed implementations, or unresolved options; “broad agreement” in Table III denotes the first category.The supplementary crosswalk maps sections to main-text elements, and Figure S1 arranges the most-voted cards by prompt with aggregated counts.Source: B. How the Day 2 board is interpreted · p. 16, Figure S1 · p. 17

S2. BIBLIOMETRIC METHODS FOR FIG. 2

  • The bibliometric survey used Web of Science records from 1991–2025, with an ARPES quantum-materials topic search yielding 7,188 records and citations binned by accrual year.The publication search excluded preprints, datasets, grants, patents, and other specified record types.Source: S2. BIBLIOMETRIC METHODS FOR FIG. 2 · p. 17
  • The survey identified 726 materials by parsing publication titles, making the count a lower bound because materials named only in abstracts are missed.Doping and strain were grouped under parent compounds, while distinct stoichiometries, end members, and films on substrates were counted separately.Source: S2. BIBLIOMETRIC METHODS FOR FIG. 2 · p. 17
  • Cross-technique publication comparisons are only indicative because search strings differ in scope, and the ARPES search curve contains 85% of records in its broader search set.The comparison uses exact-phrase searches for five probes under the same quantum-materials filter.Source: S2. BIBLIOMETRIC METHODS FOR FIG. 2 · p. 17

A. What our groups reported

  • Group self-reports suggest that published data can be a small fraction of an archive, while geometry needed to reconstruct momentum-space cuts may be missing even for recent data.The reported exabyte-per-year projection covers all techniques at DOE light sources, not ARPES specifically; reports were illustrative, not a comprehensive survey.Source: A. What our groups reported · pp. 17–18

B. Why earlier curation attempts did not persist

  • Earlier curation efforts were reported to fail amid maintenance dependence, weak depositor incentives, excessive initial scope, absent soft-metadata quality control, and separation of collection from archiving.Reuse was also hindered by difficult searches and requests to journals or authors; the proposed lesson was to standardize basic measurements and reduction first.Source: B. Why earlier curation attempts did… · p. 18

S4. CONTEXT DEPENDENCE OF ARPES DATA: A COMPACT TECHNICAL ACCOUNT

ARPES spectra require experimental context for momentum reconstruction and interpretation, so the record should keep measurements, processing, and interpretation distinct. The proposed metadata and deposition practices aim to preserve that context while remaining adaptable to community input and existing tools.

A. What momentum conversion requires

  • Momentum conversion requires recorded energy and angle context, including Fermi-level referencing, angular conventions, and sample orientation; missing context can prevent momentum placement or gap and linewidth analysis.Out-of-plane momentum also depends on photon energy and a fitted inner potential, while intensity reflects instrumental and sample conditions.Source: A. What momentum conversion requires · p. 18

B. Measurement, processing, and interpretation

  • Measurement counts and instrument state should remain distinct from processing history and model-dependent interpretation, while data quality should be judged by task rather than a universal score.The workshop supported recording measurement metadata separately from post-processing and retaining processing history in or alongside the data file.Source: B. Measurement, processing, and… · pp. 18–19

S5. PROVISIONAL METADATA SPECIFICATION AND AN EXAMPLE DEPOSITION

  • The provisional pilot specification proposes a compact mandatory core for geometry and momentum reconstruction, with extensible annotations, and is not yet a community standard.The steering committee is expected to revise it in consultation with the broader community.Source: S5. PROVISIONAL METADATA SPECIFICATION… · p. 19

A. Three metadata tiers

  • The proposed metadata tiers combine a compact structured core, optional typed extensions, and free-text context, alongside versioned schemas and explicit unknown, estimated, or not-applicable values.The workshop also stressed that adding metadata to older data should not impose excessive effort.Source: A. Three metadata tiers · pp. 19–20

B. Existing tools

  • Existing NeXus-based tools and converters provide possible starting points, but adoption of the ARPES database infrastructure described at the workshop has been slow.Participants described in-house analysis and conversion tools, and proposed metadata dictionaries with converters.Source: B. Existing tools · p. 20

C. A deposition walk-through

  • A loader can automate native-file ingest, conversion, validation, and DOI assignment, but depositors must supply missing context and help verify reproduced figures.Corrections should create a new version with a change log rather than overwrite the citable record.Source: C. A deposition walk-through · p. 20

S6. LESSONS FROM OTHER DATA ECOSYSTEMS

  • Database precedents suggest building modularly on supported infrastructure and existing practices, while recognizing that analogies to other fields have limits.The Protein Data Bank’s history was discussed alongside NexusLIMS, whose setting includes vendor-specific formats and datasets that may be abandoned over time.Source: S6. LESSONS FROM OTHER DATA ECOSYSTEMS · pp. 20–21, Table S4 · p. 22

S7. AI USE CASES, BENCHMARK DESIGN, AND VALIDATION LIMITS

The workshop reviewed AI applications and lessons from other foundation-model efforts, then described an ARPES self-energy project and proposed benchmark protocols. The account emphasizes diverse experimental data, careful validation, and limits on what current plans establish.

A. What was presented

  • The workshop catalogued ARPES AI applications by evidence status and archive needs, while noting that many published tools had not released code.An action item proposed compiling ARPES tools alongside standard computational and statistical backends for agents.Source: A. What was presented · p. 22

B. Lessons from other foundation models

  • Replication studies distinguish distribution shift from overfitting, and the ARPES questions include shifts across instruments, materials, and techniques.The authors connect this to the view that robustness depends on data breadth and quality.Source: B. Lessons from other foundation models · pp. 22–24
  • Data quality can be assessed by downstream performance rather than human judgment, supporting the workshop position that subjective judgments of “good data” should not gate deposition.The cited benchmark lesson is that human judgment can favor polished data over diversity.Source: B. Lessons from other foundation models · p. 24
  • Synthetic spectra can add scale but propagate simulator assumptions, and training on a model’s own outputs risks model collapse; their proportion and fidelity remain open questions.The workshop position was to flag simulated data and anchor validation in measured data.Source: B. Lessons from other foundation models · p. 24
  • Data-attribution methods may estimate dataset contributions, but they are computationally demanding, model-specific, and unvalidated as measures of scholarly contribution.The authors treat them as exploratory additions to traceable manifests and contributor records.Source: B. Lessons from other foundation models · p. 24

C. A worked example: self-energy prediction trained on ARPES

  • The worked project aims to predict self-energy from crystal structure and DFT bands using ARPES spectra, with 102 compounds in its workshop corpus and a target of 200 or more.Its planned test compares predictions with micro-ARPES measurements across hundreds of compositions in a material not previously measured by ARPES.Source: C. A worked example: self-energy… · p. 24

D. Benchmark protocols and the leakage problem

  • The inferred self-energy depends on the assumed DFT bare band and measurement context, while benchmark test sets must separate materials, samples, or instruments rather than spectra.The project’s initial momentum-independent self-energy assumption is expected to be unsuitable for some materials; the proposed pilot benchmarks use held-out experimental sets and conventional analysis baselines.Source: C. A worked example: self-energy… · p. 24, D. Benchmark protocols and the leakage… · pp. 24–25

S8. ACCESS, ATTRIBUTION, HOSTING, AND GOVERNANCE

Workshop discussions identified conditions for contributing ARPES data and unresolved choices around access, credit, misuse, and quality. Participants favored attribution and curated, governed data, while distinguishing identifiable dataset reuse from whole-corpus training.

A. Conditions for contributing

  • Proposed contribution conditions included explicit reuse and attribution policies, lower deposition barriers, and incentives such as reuse projects, funding, or publication requirements.A Day-1 talk also proposed recognizing data reuse in funding, conferences, fellowships, beamtime ratings, and publication requirements.Source: A. Conditions for contributing · p. 25
  • Day-2 proposals paired governed, curated data and intent/provenance with settled ownership, consent, and credit, while favoring project-specific collaboration and standardized material, doping, and temperature labels.Other proposals included facility data for internal training, with broader sharing through joint projects.Source: A. Conditions for contributing · p. 25
  • Participants favored starting with published data, while comments on access ranged from unrestricted deposit to notification, registration, or optional gating.Per-access consent drew mostly opposed comments, including concern that whole-corpus AI would not work under such consent.Source: A. Conditions for contributing · p. 25

B. Credit

  • Participants distinguished citation credit for identifiable reuse of a few datasets from credit for whole-corpus training, where individual accesses matter less.Proposed mechanisms for identifiable reuse included a DOI per dataset and a depositor-selected license; who receives credit and how deposits become citable remained open.Source: B. Credit · pp. 25–26

C. Misuse and quality

  • The leading misuse concern was unauthorized use by non-specialists; broad agreement favored education and reproducible analysis over approval gates, while gating suggestions remained individual positions.Already-published data are already public, narrowing the concern discussed to unpublished data.Source: C. Misuse and quality · p. 26

D. Hosting

Participants identified hosting and governance choices for the proposed archive, while the workshop also recorded goals ranging from near-term data organization to longer-term AI applications. Several aspirations still require assumptions and validation to be specified.

D. Hosting

  • The most-voted concern was maintenance, and no hosting option received strong positive support; participants proposed a storage census, several pilot sites, and designing for migration and mirroring.Comments noted that parallel national efforts may make a connecting superstructure necessary.Source: D. Hosting · pp. 26–27

E. Governance

  • Governance proposals included recurring meetings, a committee with defined roles and terms, and representation from laboratories, universities, theory, experiment, computing, and AI.The initial steering committee was elected anonymously, but its term, rotation, and wider-community confirmation remained unsettled.Source: E. Governance · pp. 27–28

A. Goals and definitions of success

  • Additional proposed applications included a self-driving laboratory selecting measurements, a foundation model predicting unmeasured regimes, and denoising low-count trARPES with diverse, realistic data.Participants also proposed causal-discovery AI linking growth, spectra, and properties, and restoring decades of beamline data.Source: A. Goals and definitions of success · p. 28
  • Workshop goals spanned extracting Hamiltonian parameters, automating basic ARPES, compact representations of large datasets, and inference for the trARPES energy–time resolution trade-off.The recap grouped the goals into a transferable foundation model, closing the acquisition loop, and new insight from pooled past data, while noting that some aspirations need validation.Source: A. Goals and definitions of success · p. 28
  • The most-voted goal was organizing group data in a FAIR-compatible searchable format, while definitions of success included overlooked physics insights and uptake beyond ARPES specialists.Other goals included using unpublished and failed data and building a corpus for training and benchmarking.Source: A. Goals and definitions of success · p. 28

B. Action items and the status-quo survey

The workshop proposed practical follow-up actions, a sequencing for archive scope, and a set of unresolved decisions for the pilot and steering committee. A status-quo survey would help estimate community data volume and compare metadata practices.

B. Action items and the status-quo survey

  • Proposed follow-up actions included auditing by database-experienced researchers, a metadata dictionary with converters, clarifying ownership, and a democratic vote on metadata format.A test server and uploading the easiest files first received seven votes, alongside suggestions to review existing repositories and literature before fixing a schema.Source: B. Action items and the status-quo… · pp. 28–29
  • A proposed community survey asks groups for data volume, facility versus laboratory time, major challenges, and representative files for comparing metadata fields.The questions are a template for the main-text storage census and would extend the existing group survey.Source: B. Action items and the status-quo… · p. 29

C. Scope and consensus statements

  • Workshop discussion treated published data as the first phase, with eventual scope extending to adequately described data in one archive using intent flags.A proposal for separate purpose-specific databases received no votes and three objections.Source: C. Scope and consensus statements · p. 29
  • Consensus cards called for another meeting, a paper describing findings and plans, a first committee, and invitations to computer scientists and PDB or AlphaFold participants.The next meeting was proposed for the following August.Source: C. Scope and consensus statements · p. 29

D. Unresolved decisions

  • Unresolved policy questions include the metadata core and schema, access and consent, journal deposition mandates, credit for whole-corpus training, commercial access, and synthetic data.The workshop left these decisions for the steering committee and pilot to resolve.Source: D. Unresolved decisions · p. 29
  • Other unresolved choices include the permanent host, steering committee tenure and rotation, and the boundary of the ARPES community for deposit, access, and voting.These questions were listed for resolution by the steering committee and pilot.Source: D. Unresolved decisions · p. 29
Loading 2610.15236v1…