Results 91 to 100 of about 55,072 (308)

LLM Critics Help Catch LLM Bugs

CoRR
Reinforcement learning from human feedback (RLHF) is fundamentally limited by the capacity of humans to correctly evaluate model output. To improve human evaluation ability and overcome that limitation this work trains "critic" models that help humans to more accurately evaluate model-written code. These critics are themselves LLMs trained with RLHF to
Nat McAleese, Rai Michael Pokorny, Juan Felipe Ceron Uribe, Evgenia Nitishinskaya, Maja Trebacz, Jan Leike   +5 more
openaire   +2 more sources

Large Language Model‐Informed Dual‐Track AI Framework for the Synergistic Design of Crack‐Free and High‐Strength Superalloys

Advanced Science, EarlyView.
This paper illustrates a knowledge‐augmented dual‐track AI framework for advanced superalloy design. First, Large Language Models translate metallurgical heuristics into explicit rules to rapidly prune a vast compositional search space. Subsequently, LLM‐distilled priors safely guide a reinforcement learning agent during autonomous process optimization,
Jian Yao, Junkang Wu, Jie Su, Fuzhu Wang, Zi Wang, Li Wang, Liming Tan, Lan Huang, Feng Liu, Yong Liu   +9 more
wiley   +1 more source

ChatGPT Relies More Heavily on Consonants Than on Vowels to Recognize Words

Journal of Cognition
Humans develop biases during language learning. For example, we rely more heavily on consonants than on vowels to identify words. Advances on artificial intelligence have allowed the development of proficient large language models that sometimes mimic ...
Juan Manuel Toro
doaj   +1 more source

The Conundrum of LLMs

Communications of the ACM
With LLMs, the goods are really good and the bads are really bad.
openaire   +1 more source

Trust & Safety of LLMs and LLMs in Trust & Safety

CoRR
11 ...
Doohee You, Dan Chon
openaire   +2 more sources

LLMs as Workers in Human-Computational Algorithms? Replicating Crowdsourcing Pipelines with LLMs

, 2023
LLMs have shown promise in replicating human-like behavior in crowdsourcing tasks that were previously thought to be exclusive to human abilities. However, current efforts focus mainly on simple atomic tasks.
Ding, Ziqi, Vaidos, Patricia, Liang, Jenny T., Albayrak, Maya, Axon, Alexis, Liu, Ryan, Deng, Wenxing, Tseng, Ying-Jui, Wu, Wei, Padmanabhan, Namrata, Milbauer, Jeremiah, Bertsch, Amanda, Sudjianto, Alexis, Wu, Zhijin, Ni, Xiaolin, Guo, Bill, Mandal, Ihita, Wu, Tongshuang, Ramkumar, Subhashini, Taylor, Jordan, Yang, Chenyang, Kuo, Tzu-Sheng, Zhu, Haiyi, Gururaja, Sireesh   +23 more
core  

Electrode‐Engineered Dual‐Mode Multifunctional Lead‐Free Perovskite Optoelectronic Memristors for Neuromorphic Computing

Advanced Electronic Materials, EarlyView.
A lead‐free perovskite memristive solar cell structure that call emulate both synaptic and neuronal functions controlled by light and electric fields depending on top electrode type. ABSTRACT Memristive devices based on halide perovskites hold strong promise to provide energy‐efficient systems for the Internet of Things (IoT); however, lead (Pb ...
Michalis Loizos, Konstantinos Rogdakis, Konstantinos Chatzimanolis, Katerina Anagnostou, Emmanuel Kymakis   +4 more
wiley   +1 more source

LLM-Vectorizer: LLM-Based Verified Loop Vectorizer

Proceedings of the 23rd ACM/IEEE International Symposium on Code Generation and Optimization
Vectorization is a powerful optimization technique that significantly boosts the performance of high performance computing applications operating on large data arrays. Despite decades of research on auto-vectorization, compilers frequently miss opportunities to vectorize code.
Jubi Taneja, Avery Laird, Cong Yan, Madan Musuvathi, Shuvendu K. Lahiri   +4 more
openaire   +2 more sources

Do LLMs Understand User Preferences? Evaluating LLMs On User Rating Prediction

, 2023
Large Language Models (LLMs) have demonstrated exceptional capabilities in generalizing to new tasks in a zero-shot or few-shot manner. However, the extent to which LLMs can comprehend user preferences based on their previous behavior remains an emerging
Mehta, Nikhil, Kang, Wang-Cheng, Chi, Ed, Cheng, Derek Zhiyuan, Ni, Jianmo, Sathiamoorthy, Maheswaran, Hong, Lichan   +6 more
core  

Emerging Memory and Device Technologies for Hardware‐Accelerated Model Training and Inference

Advanced Electronic Materials, EarlyView.
This review investigates the suitability of various emerging memory technologies as compute‐in‐memory hardware for artificial intelligence (AI) applications. Distinct requirements for training‐ and inference‐centric computing are discussed, spanning device physics, materials, and system integration.
Yoonho Cho, Dae‐won Kim, Yujin Kim, Jung‐Heum Na, Yuwon Jeong, Dong‐Yeop Lee, Shinhyun Choi   +6 more
wiley   +1 more source