Benedikt Ehinger

@BenediktEhinger@scholar.social
919 Followers
442 Following
1.5K Posts

Comp-#CogSci TT-Prof: 🧠, #vision, #eyetracking, #cognition, VR/mobile #EEG, methods. I blog & educate with intuition. I💖#statistics, design (thesis-art!), teaching & supervising

our lab mainly develops in #julialang

bloghttps://benediktehinger.de
lab-sitewww.s-ccs.de
teaching statistics, as one does.
🚲 Die #UniStuttgart tritt wieder in die Pedale!
Wir beteiligen uns am #STADTRADELN Stuttgart 2025. Vom 5. bis 25. Mai heißt es: Kilometer sammeln und ein Zeichen für nachhaltige Mobilität setzen.
👉Jetzt beim Team Universität Stuttgart mitmachen oder einem Subteam beitreten:
https://lnkd.in/gYuV7Htm
LinkedIn

This link will take you to a page that’s not on LinkedIn

#Scientists say they can calculate the cost of #oil giants’ role in #globalwarming
#Climate advocates hope this sort of model could result in court rulings that make polluters pay. The oil and gas industry contests the #science.
Collectively #greenhousegas #emissions from 111 #fossilfuel companies caused the world $28 trillion in damage from #extremeheat from 1991 to 2020, according to a paper published Wednesday in Nature.
https://www.washingtonpost.com/climate-solutions/2025/04/23/climate-attribution-damages-lawsuit/
https://archive.ph/pNql0
#climate #climatechange
Scientists say they can calculate the cost of oil giants’ role in global warming

Climate advocates hope this sort of model could result in court rulings that make polluters pay. The oil and gas industry contests the science.

The Washington Post

Großes Problem in #stuttgart seit ewigen Zeiten: Falschparken! Bauliche Verbesserungen dauern lange. Verkehrsüberwachung benötigt mehr Personal, um etwas bewirken zu können. Was tun? Antwort könnte sein: #Scancars. Diese könnten bald in Stuttgart zum Einsatz kommen. Wir erhoffen uns mehr Sicherheit auf der Straße. Was meint ihr dazu?

👉 https://www.stuttgarter-zeitung.de/inhalt.digitale-verkehrsueberwachung-in-stuttgart-scan-fahrzeuge-effektiver-im-kampf-gegen-falschparker.ef89174d-c32c-4453-8b64-d7cafd5fd8fb.html

@zweirat0711 @mastobikes_stuttgart

Digitale Verkehrsüberwachung in Stuttgart - Scan-Fahrzeuge: Effektiver im Kampf gegen Falschparker

Seit März dürfen in Baden-Württemberg Scan-Fahrzeuge für die Verkehrskontrolle eingesetzt werden. Das Kamerasystem kontrolliert ein Vielfaches dessen was Mitarbeiter des Ordnungsamtes schaffen. An der Uni Hohenheim läuft derzeit ein Testbetrieb.

Stuttgarter Zeitung

Unpopulär, aber ich freue mich sehr über Regen in Berlin und wünschte, es würde 2 Wochen durchregnen.

Genau deshalb: Deutschlandweit war es von Anfang Februar bis Mitte April noch nie seit Beginn der Auswertung 1931 so trocken wie in diesem Jahr.

Most #LLMs over-generalized scientific results beyond the original articles

...even when explicitly prompted for accuracy!

The #AI was 5x worse than humans, on average!

Newer models were the worst.🤦‍♂️

🔓 Accepted in #RoyalSociety Open #Science: https://doi.org/10.48550/arXiv.2504.00025

Generalization Bias in Large Language Model Summarization of Scientific Research

Artificial intelligence chatbots driven by large language models (LLMs) have the potential to increase public science literacy and support scientific research, as they can quickly summarize complex scientific information in accessible terms. However, when summarizing scientific texts, LLMs may omit details that limit the scope of research conclusions, leading to generalizations of results broader than warranted by the original study. We tested 10 prominent LLMs, including ChatGPT-4o, ChatGPT-4.5, DeepSeek, LLaMA 3.3 70B, and Claude 3.7 Sonnet, comparing 4900 LLM-generated summaries to their original scientific texts. Even when explicitly prompted for accuracy, most LLMs produced broader generalizations of scientific results than those in the original texts, with DeepSeek, ChatGPT-4o, and LLaMA 3.3 70B overgeneralizing in 26 to 73% of cases. In a direct comparison of LLM-generated and human-authored science summaries, LLM summaries were nearly five times more likely to contain broad generalizations (OR = 4.85, 95% CI [3.06, 7.70]). Notably, newer models tended to perform worse in generalization accuracy than earlier ones. Our results indicate a strong bias in many widely used LLMs towards overgeneralizing scientific conclusions, posing a significant risk of large-scale misinterpretations of research findings. We highlight potential mitigation strategies, including lowering LLM temperature settings and benchmarking LLMs for generalization accuracy.

arXiv.org

We are live!

Introducing the "Journal of Robustness Reports" – a Diamond Open-Access journal dedicated to publishing short reanalyses of empirical findings.

Check out our website and blog post about the journal:
🌐 https://scipost.org/JRobustRep
📄 https://www.bayesianspectacles.org/introducing-the-diamond-open-access-journal-of-robustness-reports/

SciPost: Journal of Robustness Reports

SciPost Journals

We are excited to announce that the Interdisciplinary International Computational Cognitive Science Summer School (IICCSSS) will take place from September 1st – 5th, 2025, in Darmstadt, Germany!
Join us for a week of insightful lectures, hands-on workshops, and discussions with leading experts in Computational Cognitive Science. Stay tuned for more details on applications and program highlights - registration will open soon! https://www.iiccsss.org/
IICCSSS

International Interdisciplinary Computational Cognitive Science Summer School

First Confirmed Footage of a #ColossalSquid—and it’s a Baby!
https://schmidtocean.org/first-colossal-squid-footage/

"The 30-centimeter juvenile #squid was captured on video at a depth of 600 meters by the #ROV #SuBastian. The sighting occurred on March 9 on an expedition near the #SouthSandwichIslands in the South Atlantic Ocean."

First Confirmed Footage of a Colossal Squid—and it’s a Baby! - Schmidt Ocean Institute

One hundred years after its discovery, the colossal squid has been filmed alive in its environment for the first time by a team on board Schmidt Ocean Institute’s R/V Falkor (too) in waters off the South Sandwich Islands. Video and photos available here. PALO ALTO, California–An international team of scientists and crew on board Schmidt … Continued

Schmidt Ocean Institute

In his new book, published today, Nachum Ulanovsky calls on the field to embrace naturalistic conditions and move away from overcontrolled experiments.

https://www.thetransmitter.org/systems-neuroscience/natural-neuroscience-toward-a-systems-neuroscience-of-natural-behaviors-an-excerpt/?utm_source=mastodon&utm_medium=org-social&utm_campaign=20250415-natural-neuroscience

‘Natural Neuroscience: Toward a Systems Neuroscience of Natural Behaviors,’ an excerpt

In his new book, published today, Nachum Ulanovsky calls on the field to embrace naturalistic conditions and move away from overcontrolled experiments.

The Transmitter: Neuroscience News and Perspectives