Best of NLP — April 2025

1. 1
   Article

   Collections·1y

   Train Your Own Large Language Model: A Comprehensive Course

   A new comprehensive course by freeCodeCamp teaches learners how to develop their own large language models (LLMs) from scratch. The course covers fundamental concepts, tokenization, Transformer architecture, and fine-tuning techniques like Low-Rank Adaptation (LoRA). Practical applications include working with chat data and developing models for underrepresented languages. Extensive resources provided include slides, notebooks, and code examples.

   255

   8
2. 2
   Article

   Hugging Face·1y

   The NLP Course is becoming the LLM Course!

   Hugging Face is upgrading its NLP course by renaming it to the LLM course, reflecting the latest advancements in AI. The revamped course will include new chapters on fine-tuning LLMs and building reasoning models, alongside maintaining and updating existing NLP content. The goal is to make cutting-edge research accessible and community-driven, with interactive exercises and live sessions available where beneficial.

   97
3. 3
   Article

   ML CMU·1y

   Machine Learning Blog | ML@CMU | Carnegie Mellon University

   Copilot Arena is a Visual Studio Code extension designed to evaluate large language models (LLMs) in real-world settings by collecting developer preferences during their actual workflow. The platform has gained over 11,000 users and supports numerous code completions and completion battles. It has shown insights into user preferences and how different models perform on various tasks. The evaluation highlights the importance of human feedback for performance metrics, contrasting with static benchmarks. Extensions to include more nuanced feedback mechanisms are encouraged.

   21
4. 4
   Article

   Collections·1y

   Mastering Prompt Engineering with Spring AI: Techniques and Best Practices

   Learn practical implementations of prompt engineering techniques using Spring AI, including configuring and tuning large language models (LLMs), and implementing these methods with Java code. Key configurations discussed include temperature control, token limits, and structured responses. Effective prompting techniques such as zero-shot, one-shot, few-shot, and self-consistency are covered, along with best practices for clarity and refinement.

   20

   1
5. 5
   Article

   Machine Learning News·1y

   RARE (Retrieval-Augmented Reasoning Modeling): A Scalable AI Framework for Domain-Specific Reasoning in Lightweight Language Models

   The RARE (Retrieval-Augmented Reasoning Modeling) framework aims to enhance domain-specific reasoning in lightweight language models by separating knowledge storage from reasoning development. Drawing on Bloom’s Taxonomy, it prioritizes contextual rationale over memory-heavy learning and uses external databases for domain knowledge. Experiments indicate that RARE-trained models outperform larger models like GPT-4 in healthcare-focused tasks, achieving over 20% higher accuracy on some benchmarks. This scalable approach suggests that focusing on reasoning skills and using structured, contextual learning can be more effective than simply increasing model size.

   14

   1
6. 6
   Article

   The Palindrome·1y

   Update: "Can Large Language Models replace mathematicians?" fixed!

   The post addresses issues encountered when using Substack's LaTeX blocks for mathematical formulas, which affected mobile readability. The author fixed the problem by manually rendering formulas and plans to contact Substack for improvement suggestions.

   10