arXiv:2312.11514v1 [cs.CL] 12 Dec 2023

Efficient Large Language Model Inference with Limited Memory - arXiv

[Submitted on 12 Dec 2023 (v1), last revised 30 Jul 2024 (this version, v3)] ... Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine ...

arXiv:2312.11514v1 [cs.CL] 12 Dec 2023. Page 2. DRAM. Flash Memory. 100 GB. 10 GB. CPU. GPU. ~ 1 GB/s. 1. 0. G. B. /s. (a) Bandwidth in a ...

arXiv:2312.11513v1 [cs.CR] 12 Dec 2023

Abstract. Prompt injection has emerged as a serious security threat to large language models (LLMs). At present, the current best-.

arXiv:2312.17120v1 [cs.CL] 28 Dec 2023

(2021b) in- troduced AMPS, a problem set ranging from el- ementary mathematics to multivariable calculus. (K-12 level) for pre-training purposes ...

arXiv:2312.15907v1 [cs.CL] 26 Dec 2023

CL] 26 Dec 2023. Page 2. of large language models more aligned with ... 2020, NeurIPS 2020, December 6-12, 2020, virtual. Junlong Li ...

arXiv:2312.07395v1 [cs.CV] 12 Dec 2023

arXiv:2312.07395v1 [cs.CV] 12 Dec 2023. Page 2. early temporal aggregation, may handicap the ability to pro- cess complex visual dependencies ...

Sparks of Artificial General Intelligence: Early experiments with GPT-4

Computation and Language (cs.CL); Artificial Intelligence (cs.AI) ... [v4] Wed, 12 Apr 2023 17:00:10 UTC (12,943 KB) [v5] Thu, 13 Apr 2023 ...

[2312.07553] Hijacking Context in Large Multi-modal Models - arXiv

Artificial Intelligence (cs.AI); Computation and Language (cs.CL) ... [v1] Thu, 7 Dec 2023 11:23:29 UTC (2,322 KB) [v2] Mon, 13 May 2024 10 ...

arXiv:2312.13043v1 [hep-ex] 20 Dec 2023

(Dated: December 20, 2023). We search for the e+e− → ηb(1S)ω and ... 90% CL upper limits on Born-level cross sections: σB(e+e− → ηb ...

arXiv:2311.05876v2 [cs.CL] 7 Dec 2023

A very important problem for retrieval-augmented large language models is to know the knowledge boundaries (Yin et al., 2023) of LLMs and deter-.