I was happy to see “someone” asking me questions about my paper, but when I read it carefully, it turned out to be an agent asking my arXiv endorsement for its AI-generated paper 🙃
Stop polluting arXiv please 🙏
Can LLMs truly reason with knowledge that conflicts with what they already believe?
Our paper TRACK, accepted as a Virtual Oral @eaclmeeting#EACL2026, shows the answer is often no. Even when you hand them the correct facts.
Find out how we did this ⬇️
I'll be at the virtual poster session (EMNLP Gather) for EMNLP 2025 from 7 pm to 8 pm EST on Nov. 5 (8 am to 9 am CST on Nov. 6).
Welcome if you have registered for the conference!
📜 Paper: aclanthology.org/2025.findings-…
🌐 Website: misinfo-prop.github.io
You may turn to an LLM for reasoning, but what if you are wrong in the first place?
LLMs usually propagate misinformation!
Even when instructed, LLMs fail to correct misinformation, showing accuracy drops: 📉 10.02%~72.20% (instruction) & 📉 4.30%~19.97% (thinking).
🧵 (1/6)
You may turn to an LLM for reasoning, but what if you are wrong in the first place?
LLMs usually propagate misinformation!
Even when instructed, LLMs fail to correct misinformation, showing accuracy drops: 📉 10.02%~72.20% (instruction) & 📉 4.30%~19.97% (thinking).
🧵 (1/6)
