LLMs can accept the same false claim differently depending on its tone, certainty, and grammatical form.
Small wording changes can make LLMs accept false claims, while larger and instruction-tuned models resist them more.
Models must decide whether to trust a user’s new claim or rely on facts stored during training.
EoBench tests this choice with about 66K false claims written in 19 styles across form, evidence, certainty, and tone.
The team evaluated 18 Gemma, Llama, and Qwen models, then kept cases where each model already knew the correct fact.
Commands, child-directed wording, formal language, and authority claims persuaded models most, while weak claims and counterfactuals persuaded them least.
Across Llama and Gemma, larger models followed false context less often, and instruction tuning usually reduced that behavior.
The finding shows that prompt wording can quietly change model answers, so evaluations and product safeguards must test linguistic framing directly.
---
– arxiv. org/abs/2607.18232
Title: "It's Not What You Say, It's How You Say It: Evaluating LLM Responses to Expressions of Belief"
Do you provide care for someone with dementia in Ireland?
If yes, please take part in a short research study on dementia care. Share your experience working in dementia care. The survey takes approx 10 minutes to complete.
Thank you 😊 #ABA#healthcare
https://t.co/ACXPbTYlM8
Where it all began 🧡🌐
This #WorldWideWebDay, we’re looking back at the early days of the Ubuntu website - the beginning of ShipIt, our program that shipped free Ubuntu CDs to anyone, anywhere in the world.
It was also the start of something big, a global community built on the promise of open source for everyone.
What are you building with open source today? Tell us in the comments below.
Yet another preprint!
A multireference picture of electronic excitations in vanadyl and copper tetraphenyl porphyrin molecular qubits 🔥🔥🔥
Excellent work by @arup_chemistry 💪💪💪
@ERC_Research
@MSCActions
@TCD_physics@ambercentre
https://t.co/e0V6GC4C04
Take a peek inside our QML datasets collection! 👀
Use these six new PennyLane datasets to train and evaluate quantum generative machine learning models:
🫧 Binary Blobs
🔃 Ising
💾 D-Wave
🔢 MNIST
⚖️ Scale Free Network
🧬 Genomic
Don’t miss them 🔗👇
Hamiltonian Simulation: one of the most promising quantum computing applications.🤞
Interested but wondering where to start? Our new PennyLane Codebook module will teach you how to implement Hamiltonian simulation algorithms using PennyLane.
Learn more! 🔗👇
@kunalkamra88 He is Devesh Dixit, stand up comedian
Taking BJP to cleaners 🔥
"Demonetization was done not for common man but to buy MLAs" 😂
Will they vandalize this venue too?