📚︎Deep Learning Fundamentals and Large Language Models

Transformer architecture optimization; efficient training and lightweight compression of large models; model distillation and quantization for acceleration; self-supervised and weakly supervised language pre-training; few-shot, zero-shot, and low-shot language learning; prompt engineering and parameter-efficient fine-tuning; long-context modeling and context window extension; inference acceleration and edge deployment of large models; green and low-power language models; emergent abilities and underlying mechanisms of large models; evaluation benchmarks and robustness testing for large models.

📚︎Fundamental Natural Language Processing Technologies

Word segmentation and part-of-speech tagging; syntactic and dependency parsing; discourse and pragmatic analysis; text representation and embedding learning; text mining and event extraction; sentiment analysis and opinion mining; text summarization and paraphrasing; machine translation and cross-lingual transfer; low-resource language processing; unified multilingual modeling; text correction and denoising; authorship attribution and text provenance; adversarial defense in text.

📚︎Neuro-Symbolic Integration and Language Reasoning

Logical reasoning in large models; Chain-of-Thought (CoT) reasoning optimization; neuro-symbolic hybrid reasoning frameworks; integration of knowledge graphs with large models; Retrieval-Augmented Generation (RAG); factuality verification and hallucination mitigation; causal language reasoning; multi-step complex problem solving; injection of symbolic rules into pre-trained models; commonsense reasoning and world modeling.

📚︎Multimodal Language Interaction

Understanding and generation of vision-language large models; joint speech-text modeling; video captioning and cross-modal question answering; multimodal dialogue systems; cross-modal alignment and feature fusion; multimodal content generation; multimodal retrieval; multimodal few-shot learning; vision-language embodied interaction; lightweight applications of multimodal large models.