Prépublication
· 2026 arXiv
EN
Personal email agents decide, for every message, what it is, whether it is urgent, whether to draft a reply, create a task, or ignore it. Their error rate is rarely measured: the stored labels were written by the system itself. We present Foyer, a self-hosted...
Federico Lucca
#cs.CL
Prépublication
· 2026 arXiv
EN
Objective: To develop a unified vision-language model for prompt-conditioned quantitative ultrasound measurement that addresses the mismatch between continuous measurement targets and discrete token-based numerical prediction. Methods: This article presents HA...
Dianye Huang, Jiamei Guo et al.
#cs.CV
Prépublication
· 2026 arXiv
EN
Large language model (LLM) platforms introducing advertising must decide how much capability ad revenue should fund and what consumers should pay to avoid ads. We study a monopolistic platform that jointly chooses costly response quality, advertising intensity...
Yuhang Ji, Fasheng Xu et al.
#econ.TH
Prépublication
· 2026 arXiv
EN
Despite advances in multilingual text representations, effective dense retrieval for Persian remains constrained by the limited availability of language-specific training resources and compact embedding models. We present Yalda Embedding, a 595.8-million-param...
Yalda Taheri, Mohammad Hassan Heydari et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Modern language models rely on subwording tokenization to define the discrete units of input processing, typically using methods like Byte Pair Encoding (BPE) that treat whitespace as a fixed semantic boundary. This design is inherently limited, as many meanin...
Pietro Di Nicco, Lucio La Cava et al.
#cs.CL
Prépublication
· 2026 arXiv
EN
Statement autoformalization translates mathematics from natural language into a proof assistant such as Lean 4, and models now do it at scale. As provers get stronger, the burden of correctness shifts from the proof to the statement, that is, to whether the fo...
Xinrui Li, Pawan Sasanka Ammanamanchi et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Large Language Models (LLMs) are increasingly used to optimize code for performance. This practice often assumes that improvements in execution speed naturally translate into overall lower energy consumption. We challenge this assumption empirically by measuri...
Sadra Emami, Benjamin Weigell et al.
#cs.SE
Prépublication
· 2026 arXiv
EN
While multimodal contrastive models like CLIP and CLAP successfully learn shared representations, their embeddings typically exhibit a persistent \emph{modality gap}. We propose addressing this separation through a purely geometric approach: changing the embed...
Shubham Gupta, Siva Reddy et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Reinforcement learning (RL) post-training improves the reasoning capabilities of large language models, but its cost is dominated by autoregressive rollout generation, which can exceed 70% of training time. Existing speculative-decoding methods are poorly suit...
Nilesh Prasad Pandey, Jangseon Park et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
Low-rank SVD-based compression offers a powerful strategy to reduce the computational costs of LLMs. However, existing methods face two key limitations: (i) global rank allocation, where uncalibrated error proxies fail to capture complex error propagation, and...
Moritz Thoma, Maximilian Groezinger et al.
#cs.AI
Prépublication
· 2026 arXiv
EN
In LLM-agent opinion simulations, each agent reports the probability that a claim is true after seeing its own previous probability and the probabilities reported by its neighbors, and that number becomes its state in the next round. If the number describes th...
Anouar Imel, Jin Qi
#cs.AI
Prépublication
· 2026 arXiv
EN
We present 14 optical imaging and eight optical spectroscopic epochs of interstellar comet 3I/ATLAS from Keck/LRIS and Gemini/GMOS, two near-infrared (NIR) imaging epochs from Keck/MOSFIRE, and four NIR spectroscopic epochs from Gemini/Flamingos-2 and GNIRS, o...
Bryce T. Bolin, Jeffrey Cooke et al.
#astro-ph.EP