Prépublication
· 2026 arXiv
EN
Many recent open-weight language models use n-gram embedding tables to improve performance. These tables store learned representations of short token sequences, such as pairs or triples of consecutive tokens, which the model retrieves during processing. Howeve...
Stephen Chung, Wenyu Du
#cs.CL
Prépublication
· 2026 arXiv
EN
Chain-of-thought (CoT) prompting elicits step-by-step rationales that appear to explain model predictions, but there is no guarantee these steps causally drive the final answer. In standard autoregressive generation, dense attention allows each reasoning step...
Gabriele Dominici, Teague McMillan et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
Multilingual pipelines such as FineWeb2 apply one recipe across a thousand languages, leaving the failure modes of individual languages unanticipated. We use an LLM once, as a corpus debugger rather than a classifier: a tool-using agent explores the corpus thr...
Marek Šuppa, Ivan Vykopal et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Instrumental convergence is often invoked as a general reason to expect capable agents to acquire or express power-seeking subgoals: self-preservation, resource acquisition, deception, or resistance to oversight which poses an important threat to the society....
Barış Alparslan, Sandy Tanwisuth et al.
#cs.LG
Prépublication
· 2026 arXiv
EN
While monolingual language model (LM) surprisal is widely established as a predictor of reading times, it cannot capture the non-native (L2) reading dynamics that are shaped by the reader's first language (L1). In this work, we investigate whether the cross-li...
Abderrahmane Issam, Irene E. Winther et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Purpose: To describe an evidence-aware retrieval evaluation protocol for a clinical retrieval-augmented generation (RAG) system over breast lesion management literature, addressing the dependence of retrieval scores on annotation coverage. Methods: An 18-docum...
Miguel Mochizuki Silva, Samuel Ismael et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Control evaluations measure whether safeguards can prevent an AI model from causing undesired outcomes even when it is intentionally trying to do so. These evaluations typically use purpose-built control settings and extrapolate their results to a model's depl...
Jeffrey Olmo, David Zhang et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Institutions that deploy large language models in regulated settings often need technical evidence of responsible AI practices before approving a deployment. COMPL-AI provides a technical interpretation of the EU AI Act that organises this evidence under six p...
Robin Quériaux, Solveig Bachellery et al.
#cs.SE
Prépublication
· 2026 arXiv
EN
Accuracy on medical examinations is common evidence offered for the clinical use of language models, yet a health system that intends to run an open model on its own hardware needs to know considerably more: whether the model recognizes when it is likely to be...
Miguel Fernandez-de-Retana, Unai Zulaika et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Research on sycophancy asks whether large language models (LLMs) bend towards what their users want to hear. We ask the reverse question: when an LLM is confident, consistent, and wrong, do users bend towards it? In a within-subjects experiment, 32 participant...
Deepthika S, Pawni Yadav et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improve the speed--accuracy trade-off, yet differ in supervision, inputs, and capacity, obscuring which ch...
Jinya Sakurai, Xun Xu et al.
#cs.AI
Texte transcrit
Wikisource
FR
Extrait : GÉOGRAPHIE ANCIENNE ET MODERNE : Géographie comparée, faisant connaître l’état et les noms divers de chaque pays aux différentes époques ; Géographie
Wikisource
#Géographie