Approfondimento sui Grandi Modelli Linguistici
Questo corso offre un'introduzione completa e approfondita alla storia dello sviluppo dei grandi modelli linguistici (LLM), alle loro architetture tecniche fondamentali, ai paradigmi di addestramento (pre-addestramento, fine-tuning e allineamento), alle estensioni multimodali, all'ingegneria dei prompt, al ragionamento a catena di pensiero, agli agenti, nonché a temi all'avanguardia come la sicurezza del modello e la protezione della privacy.
Panoramica del corso
📚 Riepilogo del contenuto
Questo corso offre un'introduzione completa e approfondita all'evoluzione dei Large Language Models (LLM), alle architetture tecniche fondamentali, ai paradigmi di addestramento (pre-addestramento, fine-tuning e allineamento), alle estensioni multimodali, all'ingegneria dei prompt, al Chain of Thought (CoT), agli agenti, nonché a temi avanzati come la sicurezza dei modelli e la protezione della privacy.
Analisi approfondita dell'evoluzione tecnologica e dell'allineamento alla sicurezza dei grandi modelli full-stack, dal pre-addestramento agli agenti generali.
🎯 Obiettivi didattici
- Distinguere tra architetture dei modelli: Identificare le differenze strutturali e i casi d'uso per modelli Encoder-only, Decoder-only e Encoder-Decoder.
- Spiegare il flusso di addestramento dei LLM: Descrivere la transizione dal pre-addestramento self-supervised al Supervised Fine-Tuning (SFT) e al Reinforcement Learning from Human Feedback (RLHF).
- Analizzare la scalabilità e il comportamento del modello: Spiegare i concetti delle Scaling Laws, delle Abilità Emergenti (apprendimento in contesto, Chain of Thought) e del fenomeno delle Allucinazioni.
- Analizzare le differenze strutturali tra architetture Encoder-only (BERT), Decoder-only (GPT) e Encoder-Decoder (T5).
- Spiegare il processo di addestramento in tre fasi: Pre-addestramento (modello base), Instruction Tuning (SFT) e Allineamento (RLHF/PPO).
- Confrontare prestazioni, leggi di scalabilità e innovazioni architetturali dei principali LLM come GPT, Llama, Qwen e DeepSeek.
- Implementare strategie di prompting zero-shot e few-shot per l'estrazione e la classificazione di dati strutturati.
- Calibrare i iperparametri del modello (Temperature, Top P, Penalties) per bilanciare output creativi e deterministici.
- Costruire prompt efficaci di Chain-of-Thought (CoT) utilizzando metodi manuali, automatici e zero-shot ("Pensiamo passo dopo passo").
- Analizzare e confrontare varianti di CoT: Differenziare tra Self-Consistency, Program of Thought (PoT), Tree-of-Thought (ToT) e Graph-of-Thought (GoT).
Lezioni 共 8 课时 · 预计 24.0h
Lezioni
Lesson
This lesson explores the paradigm shift from task-specific AI to general-purpose Large Language Models, highlighting how scaling laws and the Transformer architecture enable emergent reasoning capabilities. Students will learn how the self-attention mechanism allows models to process data in parallel and why decoder-only architectures have become the standard for modern generative AI.
This lesson explores the evolution of Transformer architectures, highlighting why the industry has shifted toward Decoder-only models due to their superior scaling laws and generative capabilities. It also examines the foundational training pipeline, covering the transition from pre-training and Chinchilla optimality to modern instruction tuning and alignment techniques.
This lesson explores the evolution of prompting strategies, ranging from zero-shot instructions to few-shot demonstrations and Chain-of-Thought reasoning for complex logic. Students will learn how to improve model performance and reliability by using structural constraints, logical decomposition, and precise formatting to ensure outputs are suitable for programmatic use.
This lesson explores the evolution of reasoning in large language models, moving from linear Chain-of-Thought to advanced structured architectures like Tree-of-Thought and Graph-of-Thought. Students will learn how these frameworks, alongside techniques like Program of Thought and knowledge editing, enable models to perform deliberate planning, verification, and complex mathematical problem-solving.
This lesson explores the challenges of AI safety, focusing on the training paradox between model helpfulness and harmlessness, as well as the mechanics of adversarial jailbreak attacks. It also introduces statistical watermarking techniques, such as the KGW framework, which use vocabulary partitioning and logit bias to embed detectable, invisible signatures into AI-generated content.
This lesson explores the fundamental conflict in AI alignment between maintaining model helpfulness and ensuring harmlessness, highlighting how jailbreak attacks exploit this tension through role-playing and cognitive coercion. Students will learn how these vulnerabilities function and examine technical strategies, such as intent analysis and robust system prompting, to defend against prompt-based bypass attempts.
This lesson explores the evolution of Multi-modal Large Language Models (MLLMs) from vision-centric systems to integrated architectures capable of processing diverse sensory inputs like audio and 3D data. Students will learn how specialized encoders and projection bridges align non-textual signals into a unified semantic space, enabling the model to effectively reason across multiple modalities.
This lesson explores the development of autonomous GUI agents that use a tripartite architecture—planning, decision-making, and reflection—to interact with software interfaces. It further examines how Reinforcement Learning and RLHF enable these agents to adapt to dynamic environments while maintaining safety and reliability against potential risks.