#NeurIPS paper by Merlijn Krale, Eline Bovy, Maris Galesloot, Thiago Simão, and Nils Jansen: On Evaluating Policies for Robust Partially Observable Markov Decision Processes #POMDPs

https://repository.ubn.ru.nl/bitstream/handle/2066/324616/324616.pdf

#TIL about #GPTZero that has built a hallucination check tool which found 100 hallucinated references in #NeurIPS 2025 accepted papers https://gptzero.me/news/neurips/

... and NeurIPS organizers just brushed it off! (sorry for the substack link) https://stevensalzberg.substack.com/p/an-embarrassing-scandal-for-ai-research

#AI #AIslop

Google’s new PaperBanana system strings together five AI agents to automatically draft scientific diagrams—right down to Matplotlib‑style plots. It works well, but still drops missing icons, a reminder that AI‑generated figures need human polish. See how this open‑source‑friendly approach performed at NeurIPS and what it means for academic publishing. #PaperBanana #AIAgents #DiagramGeneration #NeurIPS

🔗 https://aidailypost.com/news/googles-paperbanana-uses-five-ai-agents-auto-generate-diagrams

You know, I thought it would be the clankers (LLMs, gAI) turning out more and more bogus information and ultimately eating their tails. But this article as about humans, under pressure to advance in hierarchical society, are turning out the slop that is clogging our knowledge bases using clankers as a tool.

#AI
#NeurIPS
#GIGO
#LLM
#enshittification
#PublishOrPerish

https://www.theguardian.com/technology/2025/dec/06/ai-research-papers

alphaXiv (@askalphaxiv)

NeurIPS에서 열린 Latent Space 팟캐스트에서 @swyx와 대화한 내용 요약. 발언자는 댓글 기반에서 심층 연구와 ML 샌드박스 환경으로의 진화, 그리고 alphaXiv를 AI 연구용 GitHub으로 만들겠다는 비전을 설명함. 팟캐스트 링크로 자세한 이야기와 향후 로드맵 확인 가능.

https://x.com/askalphaxiv/status/2016941098699280474

#podcast #neurips #alphaxiv #ai #research

Interesting opinion piece from andrew gelmans blog (Jessica Hullman wrote it) regarding use of LLMs in academic publication. My position is essentially the same as his:

I don't care if you use LLMs do write, but I care a lot whether you accept full responsibility for the truth in both your findings and the argumentation to get there. Using LLMs and letting their hallucinations slip through indicates you do not care about truth.

https://statmodeling.stat.columbia.edu/2026/01/26/machine-learning-research-is-not-serious-research-and-therefore-hallucinated-references-are-not-necessarily-a-big-deal-agrees-a-prestigious-group-of-machine-learning-researchers/

#neurips #ai #hallucinations #academia

Peer Review Missed Fake AI Citations in Conference Papers

#ArtificialIntelligence #AcademicPublishing #NeurIPS #ResearchIntegrity

Chinese-born AI scholars in the US are building bridges at NeurIPS, linking academic networks with peers back home. Their cross‑border collaborations are boosting machine‑learning breakthroughs and tech transfer, deepening US‑China research ties. Discover how these networks are reshaping the field. #AIResearch #USChinaCollab #NeurIPS #CrossBorderResearch

🔗 https://aidailypost.com/news/chinese-born-ai-scholars-us-forge-ties-deepening-us-china

mark bissell (@MarkMBissell)

2025년은 AI 전반, 특히 해석가능성(interp) 분야에서 큰 전환의 해였으며, 작성자는 NeurIPS에서 Latent Space 팟캐스트(@latentspacepod)와의 대화에서 몇 가지 주요 테마(해석성 등)를 다룬 점을 언급하고 있습니다.

https://x.com/MarkMBissell/status/2011147204623528433

#ai #interpretability #neurips #podcast

Ashvin Nair (@ashvinair)

NeurIPS에서 @swyx와 만나 강화학습(RL)의 과거와 현재 연구에 대해 대화했다는 간단한 언급입니다. 행사에서의 RL 연구 논의가 있었음을 알리는 트윗으로, 최신 연구 동향이나 토론이 이루어진 자리였음을 시사합니다.

https://x.com/ashvinair/status/2009061569548959902

#reinforcementlearning #neurips #research