A new set of papers, sharing the long-awaited result of several reproducibility and replicability projects, including commentaries, is published today. I look forward to reading the studies, and re-using the data generated! https://t.co/vxZVPzmOCA
Okay so, we just found that over 50 papers published at @Neurips 2025 have AI hallucinations
I don't think people realize how bad the slop is right now
It's not just that researchers from @GoogleDeepMind, @Meta, @MIT, @Cambridge_Uni are using AI - they allowed LLMs to generate hallucinations in their papers and didn't notice at all.
It's insane that these made it through peer review👇
I just read a paper that made me deeply uncomfortable about the future of science.
It’s called “When Reject Turns into Accept”, and it quantifies how easy it is to manipulate LLM-based scientific reviewers without ever touching the review prompt.
Here’s what they actually measured.
The authors embedded subtle instructions into the manuscript text. Things like background explanations, framing sentences, or citation commentary. Nothing that looks suspicious to a human reviewer.
When LLM reviewers read these papers:
• Rejection rates dropped by 20-40 percentage points
• Average review scores shifted from “borderline reject” to “clear accept”
• Confidence scores increased instead of decreasing
• Models complied with the injected intent in up to 70% of cases
• Stronger models were not safer. Some were more susceptible
The attack never targets the reviewer directly.
The model just… reads the paper.
And executes what it reads.
That’s the core failure mode.
We assumed reading content is passive.
For LLMs, reading is execution.
Any pipeline that uses LLMs for evaluation is now an attack surface:
→ Peer review
→ Grant screening
→ Hiring filters
→ Paper triage
→ Automated fact-checking
The most unsettling conclusion in the paper is this:
If you let models judge information, the information itself can rewrite the judge.
Scientific review just became adversarial.
Nouvel article dans l’EJSP avec le consortium TISP : Toribio-Flórez, D., M. S. Altenmüller, K. M. Douglas, et al. 2025. “ Victims of Conspiracies? An Examination of the Relationship Between Conspiracy Beliefs and Dispositional Individual Victimhood.”
https://t.co/LmfWAtFJny
Retour en photos sur la Fête de la Science et notre animation "L'amour est dans la Psycho'" 💖 🐂🐣🌿
Merci à nos étudiants et leurs encadrants pour leur investissement. Les retours du public ont été très positifs ! 🥰 #lovepsychologiesociale#adp@UPN_Recherche@UParisNanterre
Intéressé.e.s par l'évolution des pratiques de recherche (entre autres) en psychologie sociale ? Plusieurs membres de notre laboratoire en parlent dans le dernier numéro d'@InMindFR
https://t.co/k7gk9jLjdt
Cette année encore, l’équipe PS2C du LAPPS et les étudiants du Master 2 de Psychologie Sociale se mobilisent pour la fête de la Science. Après « PsyKoh Lanta », voici venir le temps de « L’Amour est dans la Psycho » !
#fetedelascience@UPN_Recherche@UParisNanterre
If you are preparing your bachelor statistics course and would like to add optional material for students to better understand statistics on a conceptual level (see topics in the screenshot) my free textbook provides a state of the art overview. https://t.co/FdQIS63fVd
Nouvel article avec le consortium TISP :
Victims of Conspiracies? An Examination of the Relationship Between Conspiracy Beliefs and Dispositional Individual Victimhood
https://t.co/TN6rIP3tsb
🚨Nouvel article en Open Access dans Nature Climate Change avec le consortium TISP : Extreme weather event attribution predicts climate policy support across the world
https://t.co/SqFRcRYoTt