Science AI Journal
Decision Sciences · Research topic

Open research questions in Scientific Computing and Data Management

184 unresolved questions extracted from the limitations and future-work sections of 557 Scientific Computing and Data Management papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • Existing solutions decouple data and model provenance. Existing solutions provide non-tamper-evident security. Existing solutions use trust assumptions for central entities. Existing solutions are unscalable for IoT-sized data.

    Tamper-Evident Data and Model Provenance for IoT-Based Machine Learning Using Blockchain and Off-Chain Storage · 2026 · DOI

  • Can we trust agentic reviewers for agent-generated papers? SAR and PR both over-credit agent- generated papers relative to human reviewers (§5.1). Future automated reviewers should combine with principled calibration against human review. Faithfulness over complex tasks. Frontier model providers increasingly advertise faithfulness as a core capability of their agents, yet under our open-ended end-to-end research setting we still observe substantial fabrication (§5.2). The claim of faithful behaviour does not yet survive contact with sufficiently complex tasks. Future work should focus on training agents to be faithful end-to-end rather than only on individual reasoning traces. Better experiment-planning agents and scaffolds. The major challenge for auto-research is experimental rigor (§5.2). Closing this gap will require improvements: stronger agent capabilities and scaffolds that harness agents for designing and executing rigorous experiments end-to-end.

    How Far Are We From True Auto-Research? · 2026

  • Further research is needed to address the limitations of ProHunter, - Future work could focus on improving the accuracy and efficiency of ProHunter, - Research on integrating ProHunter with other threat hunting solutions could be beneficial

    ProHunter: A comprehensive APT hunting system based on whole-system provenance · 2026 · DOI

  • Significant memory and time overhead due to the extremely large provenance graphs. Imprecise segmentation of APT activities from provenance graphs due to their intricate entanglement with benign operations. Poor alignment of attack representations between CTI-derived query graphs and provenance graphs due to their substantial semantic gaps.

    ProHunter: A comprehensive APT hunting system based on whole-system provenance · 2026 · DOI

  • The lack of guidance on designing efficient, scalable, and reproducible AI-driven HPC workflows. The need for a framework to transition from rigid execution pipelines to adaptive, intelligent computational environments. The challenge of managing heterogeneous resources and complex workflow orchestration in AI-driven HPC workflows.

    Twelve quick tips for designing AI-driven HPC workflows · 2026 · DOI

  • Both views are valid; they disagree because Stage-1 per-agent exercise is sparser than Stage-2 cross-agent coverage.

    DecisionBench: A Benchmark for Emergent Delegation in Long-Horizon Agentic Workflows · 2026

  • Synthetic perturbations appear to offer inexpensive calibration data for LLM evaluators in biomedical ML, where expert review is scarce.

    Four Ledgers, Not One Score: Responsible Communication of LLM-Judge Calibration in Biomedical ML · 2026

  • Environment specification and full rerunnability remain weak for every evaluated system, ours included, so reproducibility is still an open problem for scientific agents.

    OpenAI4S: Code as Action, Science as Sessions · 2026

  • These assumptions are often implicit and rarely examined, even though results can depend strongly on the synthetic model being used.

    Cophylogeny simulators are not interchangeable: similarities, differences and structural biases in synthetic host-symbiont · 2026 · DOI

  • The determinism claim is untested: the specification claims two architects given one rule set produce one assignment, and that has not been tested with two architects — a procedure can be deterministic on paper and still admit judgement where a workload is assigned to a class.

    Cross-Border AI Architecture Patterns: Three Patterns, and Why the Fourth Is Not a Choice · 2026 · DOI

  • The determinism claim is untested: the specification claims two architects given one rule set produce one assignment, and that has not been tested with two architects — a procedure can be deterministic on paper and still admit judgement where a workload is assigned to a class.

    Cross-Border AI Architecture Patterns: Three Patterns, and Why the Fourth Is Not a Choice · 2026 · DOI

  • Section 13 is explicit about what a reference architecture does not establish — there is no implementation and no evaluation here — and about the open problems that remain.

    The Knowledge Layer: A Reference Architecture for Delegated AI Action in Regulated Institutions · 2026 · DOI

  • Despite their potential influence, the specific contributions of STEM professionals in AI policy remain underexplored.

    Developing a Typology of Roles for STEM-Trained Professionals in AI Policy Engagement · 2026 · DOI

  • The paper concludes that future research should focus on improving model efficiency, enabling multimodal data fusion, and enhancing compatibility with existing industry tools to realize the full potential of LLMs and support the digital transformation of the AECO sector.

    Application of Large Language Models in the AECO Industry: Core Technologies, Application Scenarios, and Research Challenges · 2025 · DOI

  • However, our laboratory routines and our scientific practice of communicating scientific results are insufficient to ensure the reproducibility and scalability of experiments, and data management has become a bottleneck to progress in biocatalysis.

    FAIR Data and Software: Improving Efficiency and Quality of Biocatalytic Science · 2024 · DOI

  • The need for a comprehensive testing and validation framework for EOSC Beyond Pilot Nodes. The gap in establishing a federated European Open Science Cloud infrastructure.

    EOSC Beyond D16.1 Plans for the testing and validation of the EOSC Pilot Nodes · 2026 · DOI

  • HiRSE Seminar | Metadata for Research Software | M.Gruenpeter | 25/03/2026 | CC-BY 4.0 | #25/46 Objective: #RSMD_cheklist To ensure the collection, curation, and maintenance of research software metadata, the following general requirements are recommended for end users, including researchers, software engineers, curators, and institution staff.

    Making Research Software Visible, Citable, and Preserved: A Metadata Deep Dive for RSEs · 2026 · DOI

  • Inappropriate or superficial use of containers can undermine their benefits. There is a need for practical tips on using software containers effectively in computational biology research.

    Nine quick tips for software containerization · 2026 · DOI

  • The complexity of writing parallel applications is a significant challenge. The need for a user-friendly interface for managing distributed, task-based workflows is identified.

    DistributedWorkflows.jl - A Julia interface to a task-based workflow management system. · 2026 · DOI

  • This paper and the DistributedWorkflows.jl package intro- duce a user-friendly interface for managing distributed, task-based workflows. It enables users to generate, visualize, compile, and launch workflows represented as Petri nets through simple meth- ods. The package provides a fully documented public API, allow- ing users to locally test applications before deploying them on ex- pensive clusters, ensuring cost efficiency. With binaries available for multiple Linux distributions, the tool is currently best suited for long-running processes. Upcoming updates aim to enhance the package with specialized transition types for reduced boilerplate code, additional workflow examples to guide users, and convenience functions that streamline Fig. 6. Part of a Petri net modeling different repair synthesis pathways of DNA as in [26]. the process of creating and managing workflows. Furthermore, im- provements to the user interface are planned, making the tool even more intuitive. These features are expected to expand the package’s utility, enabling broader adoption and more efficient handling of distributed workflows across various applications. 8. ACKNOWLEDGEMENTS We extend our gratitude to Fraunhofer ITWM for funding the Ph.D. studies of the first author. As well as, enabling her to work on and develop DistributedWorkflows.jl. We are also deeply grate- ful to Prof. Dr. Anne Frühbis-Krüger and Prof. Dr. Claus Fieker for their invaluable insights into the needs of domain scientists, which guided the design of this package with a strong focus on user experience. Our sincere thanks goes to Dr. Tiberiu Rotaru for his expertise in GPI-Space, particularly during the initial stages of integrating our Julia prototype with C++.

    DistributedWorkflows.jl - A Julia interface to a task-based workflow management system. · 2026 · DOI

  • The widespread over-allocation of resources leads to increased queue wait times, costs, and carbon emissions. There is a need for automatic right-sizing and efficient scheduling of job resource requests within a science gateway.

    Right-Sizing Compute Resource Allocations for Bioinformatics Tools with Total Perspective Vortex · 2026 · DOI

  • Looking ahead, two directions stand out. First, the automation of the shared database lifecycle remains a key priority. While defaults curated by experienced administrators and validated against production data have proven highly reliable, scaling to thousands of tools will require semi-automated updates. Incremental approaches, such as bracketing input sizes for tools with predictable scaling or selectively benchmarking high-impact applications, could provide efficient pathways toward automation without overreliance on machine learning. Second, there is potential for closer integration with live system telemetry. The meta-scheduling function at Galaxy Australia already queries backend load in real time, but similar approaches could extend to memory or I/O characteristics. Combining static defaults with live data would yield more adaptive scheduling while maintaining interpretability. Third, the shared TPV configuration database, combined with historical job metrics, could serve as training data for ML models that infer baseline configurations for new tools lacking curated entries. This is particularly appealing for the cold-start scenario where a new tool is added and no prior execution data exists. Such a hybrid approach could complement the rule-based system by providing informed starting points while retaining TPV’s interpretability and administrator override capabilities.

    Right-Sizing Compute Resource Allocations for Bioinformatics Tools with Total Perspective Vortex · 2026 · DOI

  • The paper does not provide a comprehensive comparison with all existing early exiting methods. The evaluation is limited to twelve datasets and eight foundation models.

    EPEE: towards efficient and effective foundation models in biomedicine · 2026 · DOI

  • Evaluating EPEE on more datasets and foundation models. Investigating the application of EPEE in other domains beyond biomedicine. Developing more efficient and effective early exiting methods.

    EPEE: towards efficient and effective foundation models in biomedicine · 2026 · DOI

  • Despite this expansion, there exists no standardized, cryptographically verifiable infrastructure capable of proving the provenance, reproducibility, model origin, training lineage, or verification status of AI-generated artifacts.

    INTERGET v2.0: The Provenance and Autonomous Coordination Layer of the Sovereign Web — Intelligent Network for Trusted Execution of Recorded Generative Experiences and Trajectories · 2026 · DOI

Most-cited papers in Scientific Computing and Data Management

Most recent work

Find a gap in your own Scientific Computing and Data Management sub-topic

This page shows what the Scientific Computing and Data Management literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Decision Sciences

Psychometric Methodologies and Testing
1,430 open questions
scientometrics and bibliometrics research
1,033 open questions
Evaluation and Performance Assessment
730 open questions
Technology Adoption and User Behaviour
537 open questions
Decision-Making and Behavioral Economics
484 open questions
Ethics in Business and Education
403 open questions
Meta-analysis and systematic reviews
313 open questions
Construction Project Management and Performance
301 open questions
184 open questions have been extracted from the limitations and future-work passages of 557 Scientific Computing and Data Management papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.