Nils Durner's Blog Ahas, Breadcrumbs, Coding Epiphanies

LLM Benchmarks: Progress and Gaps

Ethan Mollick, Associate Professor at The Wharton School, recently noted some significant gaps in current LLM benchmarking: No benchmark for LLM hallucination rates Few benchmarks with human comparisons Lack of common benchmarks for use cases like innovation, writing, persuasion, human interaction, education, and creativity Mollick poi... Read more

[UPDATED] Gpt 4o

description: “Introduces GPT-4o, highlighting architecture improvements, performance gains over GPT-4, and multi-modal input capabilities.” layout: post title: “Updated: GPT-4o” date: 2024-05-14 last_updated: 2024-05-14 tags: [gpt-4, chatgpt, summarization] — (Updated on Jul 18) So OpenAI have (pre-)released a new member of the GPT-4 family. T... Read more

Italian LLM Benchmark: INVALSI for AI

The University of Milano-Bicocca has published a significant work for Generative AI in Italy. As Alessandro Vitale notes in his LinkedIn post, there was previously no benchmark to understand how well LLMs performed in Italian. The new benchmark adapts INVALSI tests, which are typically given to Italian students in elementary, middle, and high sc... Read more

LivePortrait: Animating the Static

A video that’s currently captivating my social media timeline demonstrates a fascinating leap in AI-driven animation. Developed by Chinese research groups, this demo represents a significant milestone in what I’d love to see in a “Generative AI” product or service. The technology, called LivePortrait, animates static images based on a driver v... Read more

Improving LLM User Interfaces: The Case for Conversation Forking

Maxime Labonne, Staff Machine Learning Scientist at Liquid AI, recently posited that while the models themselves have made significant progress, user interfaces haven’t kept pace. Labonne points out that current LLM interfaces don’t align well with how people typically use these models. Users often engage in back-and-forth conversations, edit p... Read more