#Scienceouverte
📢Conférence "Le projet Software Heritage et la gestion de vos logiciels de recherche"💡
Avec Simon Delamare, ambassadeur @swheritage et ing. de recherche au laboratoire LIP
📅 2 juin 11h @ENSdeLyon & online
🖊️S'inscrire ➡️ https://www.bibliotheque-diderot.fr/acces-rapides/formulaires-de-contact/inscription-au-printemps-de-la-donnee-2026-ens-de-lyon
#openscience #researchsoftwares #opensourcce #ENSLyon #reproducibility #reproductibilité #codes #logiciels #softwares #digitalpreservation #PDLD2026
Is it really bad that only 50% of social science papers are reproducible? - LSE Impact

Three papers in Nature find that around half of social science studies hold up under replication, reproducibility, and robustness tests. Is this a bad thing

LSE Impact - Understanding impact and practice in academic research

Я хотел повторить Growing Neural CA за вечер. Ушёл месяц

Месяц назад я прочитал на Хабре статью про нейронные клеточные автоматы. Маленькие нейросети управляют клетками на сетке, клетки сами собираются в букву T или крест, и всё это обучается без учителя через что‑то вроде эволюции. Я подумал: круто, повторю за пару вечеров, посмотрю, как себя ведёт. Эта статья — о том, что было дальше. Спойлер: пара вечеров превратилась в месяц, я провёл 22 эксперимента, упёрся в потолок IoU 0.44 на простой букве T, и главное, чему научился — вообще не о нейросетях.

https://habr.com/ru/articles/1039694/

#neural_cellular_automata #neuroevolution #genetic_algorithms #neural_networks #claude_code #ml_engineering #research #pytorch #optuna #reproducibility

Я хотел повторить Growing Neural CA за вечер. Ушёл месяц

22 эксперимента, 9 потолков, один champion и неприятная правда про дисциплину эксперимента Месяц назад я прочитал на Хабре статью про нейронные клеточные автоматы. Маленькие нейросети управляют...

Хабр

I am excited that my submission at JOSE (Journal of Open Science Education; https://jose.theoj.org/) has now been assigned an editor and a first reviewer! 🥳

🚨 We are still looking for a second reviewer: https://github.com/openjournals/jose-reviews/issues/347! 🚨

Would you like to review a tutorial on #Quarto for research (with a special focus on #humanities researchers with no programming background)? If so, sign up as a reviewer here: https://docs.google.com/forms/d/e/1FAIpQLScll1kTfY3tcdL23TIv1pWXKuXUcnwrXk_bIl5fkLQnGbqb7A/viewform! This is a really different (and I personally think better!) reviewing experience than traditional academic publishing: reviews follow a checklist principle and reviewers' contributions are public.

#OpenPeerReview #OER #OpenScience #reproducibility

Journal of Open Source Education

Journal of Open Source Education (JOSE) is an educator friendly journal for publishing open-source educational materials and software.

Now available: written transcripts of our #OpenScience Bites #podcast episodes!

Prefer reading over listening, or want to quickly revisit key insights? You can now access full transcripts alongside the episodes.

🔗 https://www.rug.nl/research/openscience/podcast/ (via the drop-down menu)

#OpenScience #OpenEducation #PublicEngagement #Reproducibility #OpenResearch #Accessibility

Brilliant talk about #reproducibility by Emmanuel Boakye (AREN), during the latest #jrfrr2026.

A few key ideas and new and relevant aspects of his work as coordinator of an important network of research.

https://jrfrr-2026.sciencesconf.org/resource/page/id/3

Journées du réseau français de recherche reproductible - Sciencesconf.org

Vauban (@vauban_tech)

에이전트 스택이 원격 추론, 타사 서버 저장, 결정론적 재생 불가, 암호학적 검증 불가, trajectory 미내보내기 상태라면 사용자는 AI를 ‘소유’하는 게 아니라 빌려 쓰는 것이라는 주장이다. AI 에이전트 인프라에서 소유권·검증 가능성·재현성의 중요성을 강조한다.

https://x.com/vauban_tech/status/2057179937681854744

#agents #aiinfrastructure #sovereignai #telemetry #reproducibility

Vauban (@vauban_tech) on X

@vauban @AnthropicAI @OpenAI @GroqInc If your agent stack: ↳ phones home for inference ↳ stores chat history on someone else's box ↳ can't be replayed deterministically ↳ can't be cryptographically verified ↳ doesn't export trajectories You don't own your AI. You rent it.

X (formerly Twitter)

Vauban (@vauban_tech)

에이전트 스택이 원격 추론, 타사 서버 저장, 결정론적 재생 불가, 암호학적 검증 불가, trajectory 미내보내기 상태라면 사용자는 AI를 ‘소유’하는 게 아니라 빌려 쓰는 것이라는 주장이다. AI 에이전트 인프라에서 소유권·검증 가능성·재현성의 중요성을 강조한다.

https://x.com/vauban_tech/status/2057181030637773229

#agents #aiinfrastructure #sovereignai #telemetry #reproducibility

Vauban (@vauban_tech) on X

@vauban @AnthropicAI @OpenAI @GroqInc If your agent stack: ↳ phones home for inference ↳ stores chat history on someone else's box ↳ can't be replayed deterministically ↳ can't be cryptographically verified ↳ doesn't export trajectories You don't own your AI. You rent it.

X (formerly Twitter)

„Have you developed some tools?“

Today I had the pleasure to meet interesting people from all around the #NFDI at the 4th NFDI Berlin-Brandenburg network meeting (https://events.hifis.net/event/3380/).
The event was about #provenance and #reproducibility and people represented diverse fields, spanning astronomy, medicine, chemistry and history.

I am always impressed by the wealth of practical solutions and tools, e.g., to document workflows and create metadata. I too enjoy coding, but sticking to the critical discussion is definitely a comfort zone in sociology…

PS thanks @wikimediaDE for hosting the event. Sitting on a wooden box and having some mate is the most Berlin thing I have done recently…

New preprint: "A Framework for Reproducible AI-Assisted Research in the Social and Behavioral Sciences"
https://osf.io/preprints/psyarxiv/xdtqh_v1
#reproducibility #metascience #openscience
OSF