AI for Researchers · Visual Deck

Session 5: Data, Code & Your AI Workflow

What it can do with your data, where it fails invisibly, and the four habits that make either one defensible

{{Presenter Name}}

Landscape as of August 2026

AI for ResearchersLandscape as of August 20261 / 30

2-Minute Recap

The Series So Far


flowchart TB
  a1["S1
Capability/Failure Map
+ 2026 Tool Taxonomy"]:::done --> cap a2["S2
Literature-Discovery
Tool Matrix"]:::done --> cap a3["S3
Summary-Trust
Triage"]:::done --> cap a4["S4
Publisher AI-Policy
Comparison Table"]:::done --> cap cap(["Session 5 · today
the capstone — one page you take home"]):::today classDef done fill:#23204c,stroke:#8f8cb8,color:#b9b7d6,font-size:17px classDef today fill:#ab7d22,stroke:#d9b36c,color:#14132b,font-size:18px

CRIT — Context · Role & register · Instructions & constraints · Task, then iterate — travels with every stage.
"Generating analysis code" sits in the Unreliable zone: never the only pass, always spot-checked.

AI for ResearchersLandscape as of August 20262 / 30

Learning Objectives

By the End of This Session…


1Explain the four reproducibility guardrailskeep the code · pin the versions · set the seed · verify one number by hand
2Run an AI-assisted statistical analysis or visualizationon a sample dataset, live
3Apply a no-code analysis pathto a dataset, without writing code directly

Objectives 4–6 next · full verbatim wording in the reference deck and curriculum

AI for ResearchersLandscape as of August 20263 / 30

Learning Objectives

…And You Will Also Be Able To


4Evaluate an AI-assisted analysis before trusting itfor correctness and reproducibility
5Assemble a personal AI workflowspanning discovery, reading, writing, and analysis
6Identify strategies for staying currentas agentic coding/analysis tools continue to evolve

Full verbatim wording in the reference deck and curriculum

AI for ResearchersLandscape as of August 20264 / 30

Section 01

What AI Actually Does to Your Data

Three jobs, one mechanism — and the mechanism is identical whether or not you ever see a line of code.

AI for ResearchersLandscape as of August 20265 / 30

Literacy Foundation

Three Jobs You Are Actually Asking It to Do


Job 1Write the codeload, clean, reshape, model — the script you would otherwise have written or commissioned
Job 2Advise on the statisticswhich test, which assumptions, how to read the output
Job 3Make the figureand with it, the claim the figure implies to a reader who will never see the data

These fail in different ways — only one fails loudly. All three sit in the Unreliable zone, and Rule 1 is the entry condition: if you could not catch the error, the task is not in the Safe zone.

AI for ResearchersLandscape as of August 20266 / 30

The Mechanism, In Their Own Words

"No-Code" Is Not No Code. It Is Code You Have Not Read.


"For some data-analysis tasks, ChatGPT writes and runs Python code in a stateful Jupyter notebook environment."

OpenAI, Data analysis with ChatGPT [1]

Excel: Copilot uses "Python-based analysis…" — "expand the code to see how the answer was generated" [5]
Colab's agent shows every suggestion and waits for Accept or Accept and run [3]
but one major assistant's upload docs do not state it writes or runs visible Python at all [4]

The question is not "do I code?" — it is "can I get the code out?" A property of the tool, chosen before you upload.

AI for ResearchersLandscape as of August 20267 / 30

Before You Upload Anything

Where Does Your Dataset Go?


on by defaultuploads may train the model unless you turn it off: "Improve the model for everyone", Settings > Data Controls [1]
30 daysone vendor's execution containers "expire 30 days after creation" — sandboxes are isolated, internet-free, and ephemeral [2]
1 h / 7 daysone tool deletes free-tier files after 1 hour of inactivity; 7 days on paid plans — retention differs sharply by product and tier [6]

Their own warning: a file-handling assistant "can be tricked into sending information from its context… to malicious third parties" [2]. Excel: "runs within the compliance boundary of your organization" [5]. Your institution's data-governance policy outranks every line above — public, consented, de-identified or published data is the only safe default.

AI for ResearchersLandscape as of August 20268 / 30

The Evidence — Measured, Not Asserted

How Well Does It Actually Do an Analysis?


ScienceAgentBench superseded by AstaBench: 58.0% overall, ~3% perfect end-to-end discovery [48] · DSBench: no re-run found [13] · SciCode board stale; 60.2% via Artificial Analysis [14] [50] · CORE-Bench 95.5% manually scored; HAL paused [15] [49] · StatQA/StatABench [10] [17] — rows comparable to your expectations, not to each other [9]

AI for ResearchersLandscape as of August 20269 / 30

Critical Literacy — The Key Slide

The Error That Matters Is Not Arithmetic


"LLMs primarily make applicability errors, whereas humans mostly make statistical task confusion errors."

Zhu et al., StatQA — 11,623 items, NeurIPS 2024 [10]. An applicability error is silent: the test ran, the arithmetic is right, the assumptions did not hold — and nothing in the output says so.

68.6%GPT-5.1 on the successor benchmark, two years newer — and half its audited failures are still tool selection [17]
< 13%of expert annotators' modelling decisions reproduced by agents given the same research questions (2024 cohort) [11]

Fair counterweight: six models chose the correct test in 100% of 20 textbook scenarios — basic, free tiers, one day in August 2025 [19]. Choosing the test is close to solved. Applying it to your data is not — discrepancies cluster in post-hoc tests, confidence intervals, sign conventions, non-parametric tests [18]

AI for ResearchersLandscape as of August 202610 / 30

The Evidence

Ask It Twice. The Answer Is Not the Same.


21–81%workflow reproducibility across 15 model pairs, 1,032 tasks, three task sets (21–55% · 30–75% · 48–81%; human experts: 66.5%) [12]
480runs of the same analysis, same settings: "considerable variation" [43]
75.8%of coding tasks gave zero identical outputs across repeat requests [44]

"Setting the temperature to 0 does not guarantee determinism" [44] — the authors' own fix: run it multiple times independently and consider the distribution [43]

AI for ResearchersLandscape as of August 202611 / 30

The Evidence

And in the Code Itself, Three Documented Failures


43.53%of hallucinations in 1,380 annotated snippets were "task requirement conflicts" — it solved a different problem, without crashing [20]
29.5%of AI-generated Python snippets in real repositories carried a security weakness (43 CWE categories); static analysis + the assistant fixed up to 55.5% [22]

Stated precisely: 43.53% is a share of the hallucinations found, not of all code [20]. The one to fear: code that runs, returns a number, and answers a question you did not ask.

AI for ResearchersLandscape as of August 202612 / 30

The Evidence — Two RCTs That Disagree

How Fast It Feels Is Not How Fast It Is


55.8% fastera different RCT, one self-contained greenfield task — vendor-authored, silent on whether the code was correct [25]
less secure,
more confident
with an AI assistant, people wrote less secure code while being more likely to believe it was secure [23]

Honest caveats: n = 16 developers on codebases they knew for five years [24]; n = 47 students, 2022-era model [23]. The durable finding is the calibration gap, not the sign of the effect.

AI for ResearchersLandscape as of August 202613 / 30

Section 02

Reproducibility Guardrails

Four habits. Three of them were published in 2013. AI did not create the problem — it multiplied the volume of code flowing into it.

AI for ResearchersLandscape as of August 202614 / 30

The Evidence — The Pre-AI Baseline

Research Code Already Did Not Run


74%of 9,000+ deposited R files failed to run without error; 56% even after cleaning [26]
26%of 204 Science papers under a code policy could be reproduced — artifacts obtainable for only 44% [28]
879 of 27,271notebooks from 3,467 biomedical papers reproduced their own results [27]

The dominant cause was not bad science — it was undeclared dependencies and hard-coded file paths [27]. Of 296 deposited R projects, 98.8% carried no formal dependency description; 25.87% ran after automated environment rebuilds [45]. The question is not whether AI is worse than a human analyst — it is whether your result survives being re-run.

AI for ResearchersLandscape as of August 202615 / 30

Series Artefact

The Four Reproducibility Guardrails


1Keep the code — before you close the tabsave it beside the data and re-run it outside the chat; sandboxes "expire 30 days after creation" [2] — skip it and you have a transcript, not a result
2Pin the versions — tool, libraries, dateundeclared dependencies were the top failure cause across 27,271 notebooks [27]
3Set the seed — and say wherein the prompt and in the code; Rule 6 of the 2013 ten rules: "note underlying random seeds" [29]
4Verify one number — by hand, every timerecompute one figure outside the tool; applicability errors are invisible [10] and the same question yields a different workflow 21–81% of the time [12]

Guardrails 1–3 are Sandve et al.'s Rules 1, 3 and 6 — published in 2013 [29]. Guardrail 4 is the one AI adds: the analyst who made the choices cannot be asked why.

AI for ResearchersLandscape as of August 202616 / 30

The Rules, In Their Words

None of This Is Optional — It Is Already in the Policies


"…declare this in detail in the Methods section… The same standards for the validity and reproducibility of code apply whether the code was produced with AI assistance or not."

Elsevier, Generative AI policies for journals, updated June 2026 [42]

Springer Naturecode goes to "a repository that assigns a permanent identifier" — "providing a GitHub link only is not sufficient" [30]
COPEAI in "the collection and analysis of data" is disclosed in the Materials and Methods — how it was used and which tool [31]
ICMJEwriting assistance → acknowledgments; "data collection, analysis, or figure generation" → the methods [32]

Session 4's rule holds: the publisher page gives the shape, the journal's guide for authors gives the wording. Check it the week you submit.

AI for ResearchersLandscape as of August 202617 / 30

Section 03

If You Do Not Code

Everything so far applies to you unchanged — and one column of the next table decides whether you can obey guardrail 1 at all.

AI for ResearchersLandscape as of August 202618 / 30

Series Artefact

No-Code Analysis Paths: Can You Get the Code Out?


Assistant + code executionChatGPT, Claude
Yes — ask for it, and save it [1] [2]
no internet in the sandbox; ~50 MB CSV ceiling [1] [2]
Spreadsheet copilotCopilot in Excel
Yes — "expand the code to see how the answer was generated" [5]
results "aren't refreshable"; does not modify your workbook [5]
Notebook agentColab Data Science Agent
Yes, by designAccept, or Accept and run [3]
CSV and BigQuery sources only; runs only in your runtime [3]
Dedicated analysis toolJulius
Yes — notebooks "re-run, version, and share" [6]
free-tier files deleted after 1 h idle; size limit undocumented [6]
Point-and-click stats + AIJASP 0.98, July 2026
Partly — the file "holds analyses and interpretations" with the data [7]
AI opt-in with your own API key; data handling not documented [7]
Point-and-click stats, no AIjamovi
Yes — Syntax mode emits "equivalent R code for each analysis" [8]
sharing needs colleagues to hold the same R packages [8]

Every cell from that tool's own documentation, fetched 2026-07-28 — paths, not endorsements. New since compilation: Gemini Notebook (AI Pro) adds cloud code execution — press-reported, not vendor-documented [52] [53]; check code visibility first.

AI for ResearchersLandscape as of August 202619 / 30

Section 04

Agentic Tools, in the Present Tense

From "write me a script" to "do the analysis" — that shift already shipped: Claude Science, Gemini Notebook's code execution, Elicit's Research Agent, Codex as a general work agent. The approval prompt is the last place a human is guaranteed to be standing.

AI for ResearchersLandscape as of August 202620 / 30

The Mechanism, In Their Own Words

What "Agentic" Means, From the Vendors' Own Docs


The tool will "determine which files to make changes to, offer code changes and terminal commands for the user's approval, and iterate to remediate issues until the original task is complete."

GitHub, Copilot features — agent mode [33]

one agent runs with "strict read-only permissions by default"; unmatched commands "default to requiring manual approval" [34]
another defaults to a read-only sandbox with "network access turned off" [35]
inside a notebook: "Agents request permission before writing files, running commands, or using MCP tools" [36]

Their own warnings: "no system is completely immune to all attacks" [34]; "prompt injection can cause the agent to fetch and follow untrusted instructions" [35]. The approval prompt is not a formality. It is the control.

AI for ResearchersLandscape as of August 202621 / 30

The Evidence — The Trajectory, Re-Measured

The Trajectory Is Real. So Is the Gap.


~3%perfect completion on AstaBench's end-to-end discovery tasks — best 2026 agent [48]
68.6%where statistics still sits [17] — framing the question and carrying a discovery end to end remains your job description

Well-specified, metric-scored tasks moved 40–50 points in two years [16] [49]. Stated honestly: MLE-bench top entries self-reported, submissions paused [16]; HAL paused — 77.8% is an Opus 4.5-class agent, not the current frontier [49]; AstaBench is Ai2's own leaderboard [48]

AI for ResearchersLandscape as of August 202622 / 30

Practice

Staying Current Without Reading Marketing


Preprints, for speedarXiv cs.CL · cs.SE · stat.ML — months before peer review, and filtering nothing [37]
Proceedings, for trustNeurIPS Datasets & Benchmarks · ICLR · ICML (PMLR) · ACL Anthology — read leaderboard pause and provenance notes as data [38]
News desks, for orientationNature and Science news pages — a lead to follow, never a citation [39]
The one durable habitwhen a number impresses you, find the task set: how many tasks, which model version, how many attempts, who paid

If there is no task set, it is not a result. It is an advertisement.

AI for ResearchersLandscape as of August 202623 / 30

Section 05

The Capstone

Five sessions, five artefacts, one page. Two rows are deliberately empty — Sessions 6 and 7 fill them in.

AI for ResearchersLandscape as of August 202624 / 30

Series Artefact · The Capstone

The Personal AI Workflow Canvas


1 · FrameSession 1
the shape of the question; a weak prompt to sharpen
Tier 1 · Safe
read every line you keep — could you catch the error? · CRIT
2 · DiscoverSession 2
search terms; candidate papers; a first synthesis
Tier 2 first, 3 for breadth · Unreliable
every citation resolved in a database before it is used · Tool Matrix
3 · ReadSession 3
Q&A over your own PDFs; extraction columns
Tier 2, grounded · Unreliable
reopen the paper whenever a number or a qualifier appears · Summary-Trust Triage
4 · AnalyseSession 5 — today
analysis code; one test; one figure
Tier 1 + sandbox, or agentic · Unreliable
keep the code · pin versions · set the seed · recompute one number by hand · Four Guardrails
5 · WriteSession 4
language & structure of text you wrote; the disclosure sentence
Tier 1 · Safe as editor, Dangerous as ghostwriter
diff every touched paragraph; disclose in the journal's own format · Policy Table
6 · Cross-field orientation — you fill this row in during Session 6
7 · Your discipline's specialised tools — you fill this row in during Session 7

One page, one project, one date — rows are stages of your work, not products, so the canvas survives your tools being replaced. Fillable version: last page of the handout.

AI for ResearchersLandscape as of August 202625 / 30

Series Artefact

Filling It In: Three Rules


Rule 1Fill the check column firstif you cannot name the check, you have not decided to delegate — you have decided to hope. Session 1's Rule 1, applied to a whole project
Rule 2One project, not a careera canvas for "my research" is a wish; one for "this paper, this dataset, this journal" is a workflow you can hand to a co-author
Rule 3Date it — re-check in six monthsevery tool named in these five sessions changed during the ten weeks we ran them. The stages did not

And leave rows 6 and 7 empty — next session's homework, and the session after that. The canvas names tasks, tiers and checks; nothing on it becomes wrong when a product is renamed.

AI for ResearchersLandscape as of August 202626 / 30

Demo Preview

CRIT, Applied to an Analysis You Can Defend


flowchart LR
  U["upload
344 penguins · CC-0
public dataset"]:::n --> E["explore
missing values
per column"]:::n --> T["one test
Welch's t-test
seed 42"]:::n --> F["one chart
~thirteen minutes
in all"]:::n --> V["seven minutes
of checking

re-run outside the chat ·
recompute one group mean ·
check the source paper"]:::hot classDef n fill:#23204c,stroke:#8f8cb8,color:#eceaf8,font-size:18px classDef hot fill:#1c1a3f,stroke:#d9b36c,color:#d9b36c,font-size:18px

C "a public CC-0 dataset of 344 penguins… a researcher, not a programmer" · R "a careful research statistician… plain language" · I "show the code every time · drop nothing without telling me · Welch's t-test · seed 42 · print every library version" · T "describe this dataset and its data-quality problems" — then iterate, then verify [40] [41]

AI for ResearchersLandscape as of August 202627 / 30

References

References (1–18)


  1. OpenAI (2026). Data analysis with ChatGPT; File storage and Library in ChatGPT.help.openai.com/en/articles/8437071 — accessed 2026-07-28
  2. Anthropic (2026). Create and edit files with Claude; Code execution tool.support.claude.com · platform.claude.com/docs — accessed 2026-07-28
  3. Google (2026). Use the Data Science Agent. Colab Enterprise documentation.docs.cloud.google.com/colab/docs/use-data-science-agent — accessed 2026-07-28
  4. Google (2026). Upload & analyze files in Gemini Apps.support.google.com/gemini/answer/14903178 — accessed 2026-07-28
  5. Microsoft (2026). Get direct answers to your data analysis questions; Data security and Python in Excel.support.microsoft.com — accessed 2026-07-28
  6. Julius AI (2026). What is Julius AI?; Files; Containers.julius.ai/docs/get-started — accessed 2026-07-28
  7. JASP (2026, 2 July). Introducing JASP 0.98: Fully Integrated AI Support; Getting started.jasp-stats.org — accessed 2026-07-28
  8. jamovi (2026). Combining jamovi and R. User manual §6.docs.jamovi.org/usermanual/um_6_jamovi_and_R.html — accessed 2026-07-28
  9. Chen, Z., et al. (2025). ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery. ICLR 2025.arxiv.org/abs/2410.05080 — accessed 2026-07-28
  10. Zhu, Y., Du, S., Li, B., Luo, Y., & Tang, N. (2024). Are Large Language Models Good Statisticians? (StatQA). NeurIPS 2024, Datasets & Benchmarks.proceedings.neurips.cc — accessed 2026-07-28
  11. Gu, K., et al. (2024). BLADE: Benchmarking Language Model Agents for Data-Driven Science. Findings of EMNLP 2024.arxiv.org/abs/2408.09667 — accessed 2026-07-28
  12. Zeng, Q., Jin, C., Wang, X., Zheng, Y., & Li, Q. (2025). AIRepr: An Analyst-Inspector Framework for Evaluating Reproducibility of LLMs in Data Science. Findings of EMNLP 2025.arxiv.org/abs/2502.16395 — accessed 2026-07-28
  13. Jing, L., et al. (2025). DSBench: How Far Are Data Science Agents from Becoming Data Science Experts? ICLR 2025.arxiv.org/abs/2409.07703 — accessed 2026-07-28
  14. Tian, M., et al. (2024). SciCode: A Research Coding Benchmark Curated by Scientists. NeurIPS 2024, Datasets & Benchmarks; official leaderboard.arxiv.org/abs/2407.13168 · github.com/scicode-bench/SciCode — accessed 2026-07-28
  15. Siegel, Z. S., et al. (2024, rev. 2026). CORE-Bench: A Computational Reproducibility Agent Benchmark.arxiv.org/abs/2409.11363 — accessed 2026-07-28
  16. Chan, J. S., et al. (2025). MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering. ICLR 2025; official leaderboard.arxiv.org/abs/2410.07095 · github.com/openai/mle-bench — accessed 2026-07-28; leaderboard re-checked 2026-09-02
  17. Zhu, Y., et al. (2026). StatABench: Dataset and Framework for Evaluating Statistical Analysis Capabilities of LLMs. Preprint.arxiv.org/abs/2606.22977 — accessed 2026-07-28
  18. Shahrul, A. I., & Syed Mohamed, A. M. F. (2024). A Comparative Evaluation of SPSS and ChatGPT-4 in Statistical Analyses. Cureus 16, e72581.doi.org/10.7759/cureus.72581 — accessed 2026-07-28
AI for ResearchersLandscape as of August 202628 / 30

References

References (19–36)


  1. Shukla, M., et al. (2025). Evaluating… Large Language Models in Statistical Test Selection. Cureus 17, e94949.doi.org/10.7759/cureus.94949 — accessed 2026-07-28
  2. Zhang, Z., et al. (2025). LLM Hallucinations in Practical Code Generation. ISSTA 2025.arxiv.org/abs/2409.20550 — accessed 2026-07-28
  3. Spracklen, J., et al. (2025). We Have a Package for You! A Comprehensive Analysis of Package Hallucinations by Code Generating LLMs. USENIX Security '25.arxiv.org/abs/2406.10279 — accessed 2026-07-28
  4. Fu, Y., et al. (2025). Security Weaknesses of Copilot-Generated Code in GitHub Projects. ACM TOSEM.arxiv.org/abs/2310.02059 — accessed 2026-07-28
  5. Perry, N., Srivastava, M., Kumar, D., & Boneh, D. (2023). Do Users Write More Insecure Code with AI Assistants? ACM CCS '23.arxiv.org/abs/2211.03622 — accessed 2026-07-28
  6. Becker, J., et al. (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. METR.arxiv.org/abs/2507.09089 — accessed 2026-07-28
  7. Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023). The Impact of AI on Developer Productivity: Evidence from GitHub Copilot.arxiv.org/abs/2302.06590 — accessed 2026-07-28
  8. Trisovic, A., et al. (2022). A large-scale study on research code quality and execution. Scientific Data 9, 60.doi.org/10.1038/s41597-022-01143-6 — accessed 2026-07-28
  9. Samuel, S., & Mietchen, D. (2024). Computational reproducibility of Jupyter notebooks from biomedical publications. GigaScience 13.doi.org/10.1093/gigascience/giad113 — accessed 2026-07-28
  10. Stodden, V., Seiler, J., & Ma, Z. (2018). An empirical analysis of journal policy effectiveness for computational reproducibility. PNAS 115.doi.org/10.1073/pnas.1708290115 — accessed 2026-07-28
  11. Sandve, G. K., et al. (2013). Ten Simple Rules for Reproducible Computational Research. PLoS Comput. Biol. 9(10), e1003285.doi.org/10.1371/journal.pcbi.1003285 — accessed 2026-07-28
  12. Springer Nature (2026). Code policy.springernature.com/gp/open-science/code-policy — accessed 2026-07-28
  13. COPE Council (2023). Position statement: Authorship and AI tools.publicationethics.org/guidance/cope-position/authorship-and-ai-tools — accessed 2026-07-28
  14. ICMJE (2025). Recommendations — Defining the Role of Authors and Contributors (II.A.4).icmje.org/recommendations — accessed 2026-07-28
  15. GitHub (2026). GitHub Copilot features.docs.github.com/en/copilot/get-started/features — accessed 2026-07-28
  16. Anthropic (2026). Claude Code — Security.code.claude.com/docs/en/security — accessed 2026-07-28
  17. OpenAI (2026). Codex — Agent approvals & security.learn.chatgpt.com/docs/agent-approvals-security — accessed 2026-07-28
  18. Project Jupyter (2026). Jupyter AI v3 documentation.jupyter-ai.readthedocs.io/en/v3/ — accessed 2026-07-28
AI for ResearchersLandscape as of August 202629 / 30

References

References (37–53)


  1. arXiv (2026). Category listings: cs.CL, cs.SE, stat.ML.arxiv.org/list/cs.CL/recent · /cs.SE/recent · /stat.ML/recent — accessed 2026-07-28
  2. Official proceedings: ACL Anthology; NeurIPS Proceedings (incl. Datasets & Benchmarks Track); PMLR/ICML; ICLR.aclanthology.org · papers.nips.cc · proceedings.mlr.press · iclr.cc — accessed 2026-07-28
  3. Nature news and Science news. Publisher news desks; orientation only, never a citation.nature.com/nature/articles?type=news · science.org/news/all-news — accessed 2026-07-28
  4. Horst, A. M., et al. (2020). palmerpenguins: Palmer Archipelago (Antarctica) penguin data. R package v0.1.0. CC0; licence verified at CRAN and in LICENSE.md.doi.org/10.5281/zenodo.3960218 — accessed 2026-07-28
  5. Gorman, K. B., et al. (2014). Ecological sexual dimorphism and environmental variability within a community of Antarctic penguins (genus Pygoscelis). PLOS ONE 9(3), e90081.doi.org/10.1371/journal.pone.0090081 — accessed 2026-07-28
  6. Elsevier (2026, updated June 2026). Generative AI policies for journals — incl. the AI-assisted code FAQ.elsevier.com/about/policies-and-standards — accessed 2026-07-28
  7. Cui, J., & Alexander, R. (2026). Same Prompt, Different Outcomes: Evaluating the Reproducibility of Data Analysis by LLMs. Preprint.arxiv.org/abs/2602.14349 — accessed 2026-07-28
  8. Ouyang, S., et al. (2024). An Empirical Study of the Non-determinism of ChatGPT in Code Generation. ACM TOSEM.doi.org/10.1145/3697010 — accessed 2026-07-28
  9. Saju, L., et al. (2025). Computational Reproducibility of R Code Supplements on OSF. Preprint.arxiv.org/abs/2505.21590 — accessed 2026-07-28
  10. Baltes, S., et al. (2026). Guidelines for Empirical Studies in Software Engineering involving Large Language Models. Empirical Software Engineering (accepted).arxiv.org/abs/2508.15503 — accessed 2026-07-28
  11. Zuckerman, A. L., et al. (2026). Ten simple rules for building a collaborative coding culture. PLOS Computational Biology 22(2), e1013970.doi.org/10.1371/journal.pcbi.1013970 — accessed 2026-07-28
  12. Allen Institute for AI (2026, 30 April). AstaBench — spring 2026 update; ICLR 2026 paper. Ai2's own agentic-science leaderboard.allenai.org/blog — accessed 2026-08-21
  13. HAL — Holistic Agent Leaderboard (2026). Princeton; standardized agent harness, currently paused for new models.hal.cs.princeton.edu · arxiv.org/abs/2510.11977 — accessed 2026-08-21
  14. Artificial Analysis (2026). SciCode evaluation — 288 test-set subproblems, current models. Independent evaluator.artificialanalysis.ai/evaluations/scicode — accessed 2026-08-21
  15. Churilov, A. (2026). The Range Shrinks, the Threat Remains: Re-evaluating LLM Package Hallucinations on the 2026 Frontier-Model Cohort. Preprint.arxiv.org/abs/2605.17062 — accessed 2026-09-02
  16. Google Workspace Updates (2026, 16 July). NotebookLM is now Gemini Notebook.workspaceupdates.googleblog.com — accessed 2026-08-21
  17. 9to5Google (2026, 16 July). Gemini Notebook: cloud code execution for AI Pro. Press report; orientation only.9to5google.com — accessed 2026-08-21