arXiv:2608. 14712v1 Announce Type: cross Abstract: Each row of a transformer's attention matrix is a probability distribution over tokens, and in trained models most of that probability lands on a single \emph{sink} token, usually the first.
By Marios Papamichalis, Regina Ruane
arXiv:2605.19352v2 Announce Type: replace-cross
Abstract: Understanding how humans and artificial intelligence systems predict and plan by interacting with their environment is a fundamental challeng...
By Subba Reddy Oota, Anant Khandelwal, Khushbu Pahwa, Satya Sai Srinath Namburi, Tanmoy Chakraborty, Bapi S. Raju, Manish Gupta
arXiv:2605. 30556v2 Announce Type: replace Abstract: CORRECTION (August 2026): the central finding of this paper is not supported.
By Nils Leutenegger
The paper investigates whether language models fine‑tuned on different harmful tasks share common computational components. Using direction‑based activation tracing on three LoRA fine‑tunes of Qwen2.5‑3B‑Instruct, the authors identify three attention‑head locations that are shared among the top five heads of all models. Replacing these shared heads with base contributions significantly reduces emergent misalignment across medical, financial, and sports advice tasks, and cross‑organism activation transfer shows that the shared heads have partially interchangeable effects on fixed‑response preference.
By Muhammad Haseeb, Muhammad Tahir
arXiv:2607. 18114v1 Announce Type: cross Abstract: Modern LLMs are alarmingly susceptible to surprisingly simple immaterial changes of input prompts: a casual hint, an incorrectly labeled few-shot example, or a fake prior assistant turn often flips an originally correct answer.
By Prakhar Gupta, Terry Jingchen Zhang, Florent Draye, Bernhard Sch\"olkopf, Zhijing Jin
arXiv:2607. 01002v1 Announce Type: cross Abstract: In long-context use, large language models frequently synthesize answers from the meaning of a relevant context span rather than literally copy-pasting them.
By Aryo Pradipta Gema, Beatrice Alex, Pasquale Minervini