Reinforcement learning and bandits for speech and language processing: Tutorial, review and outlook

Lin, Baihan

doi:10.1016/j.eswa.2023.122254

Cited by 5 publications

(1 citation statement)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Incorporating feedback loops in AI systems has become an integral part of advancing their adaptability and accuracy. Initial studies in this area were rooted in basic machine learning paradigms, where systems were trained to modify their behaviors based on explicit feedback signals [34,35,36,37]. This form of learning, often seen in reinforcement learning scenarios, laid the foundation for more complex feedback mechanisms.…”

Section: Feedback Loops In Artificial Intelligencementioning

confidence: 99%

LLaMALoop: Enhancing Information Retrieval in LLaMA with Semantic Relevance Feedback Loop

Tsai,

Kuo,

Huang

2023

Preprint

View full text Add to dashboard Cite

This paper introduces LLaMALoop, an innovative enhancement to the Large Language Model (LLaMA), through the integration of a Semantic Relevance Feedback Loop (SRFL). This enhancement addresses the challenge of dynamic and context-sensitive information retrieval, a limitation in standard language models reliant on static training datasets. The SRFL enables LLaMALoop to adapt in real-time to evolving user queries, refining its comprehension and response accuracy through continuous learning from user feedback. The study's experimental setup involves rigorous testing across several semantic tasks, demonstrating significant improvements in the model's ability to process and interpret complex linguistic structures and user intents. Notable advancements in Semantic Role Labeling, Word Sense Disambiguation, Textual Entailment, Frame Semantic Parsing, and Commonsense Reasoning are presented. While the SRFL enhances semantic processing capabilities, it also introduces computational trade-offs, particularly in processing time. Qualitative analysis further highlights the model's improved user interaction and adaptability. LLaMALoop sets a new benchmark in the adaptability and responsiveness of language models, paving the way for more user-centric, context-aware AI systems. The findings significantly contribute to the field of AI and LLM research, particularly in areas focusing on dynamic learning and user-centric model adaptation.

show abstract

Section: Feedback Loops In Artificial Intelligencementioning

confidence: 99%