Tatsuro KAWAMOTO·Jul 10How We Designed a Blind Tasting with Five CoffeesMore than a typical coffee flight, less formal than cupping
Tatsuro KAWAMOTO·Jul 9Do You Really Know What Your Coffee Tastes Like?A blind tasting challenge with five specialty coffees
Tatsuro KAWAMOTO·Jul 5Recreating the Lady Tasting Tea Experiment Is Harder Than It SoundsThe hardest part was not the statistics
Tatsuro KAWAMOTO·Jul 2Can People Taste the Difference Between Milk-First and Coffee-First Café au Lait?“Lady Tasting Tea” Revisited
Tatsuro KAWAMOTO·Jul 1, 2024Multi-Agent Prompts in LLMs: Do They Really Work?Attempts to improve LLM performance range from modifying model architectures to prompt engineering such as the Chain of Thoughts (CoT). In…
Tatsuro KAWAMOTO·Jun 17, 2024Can ChatGPT Detect Hate Speech?Ethical issues have been a concern since the very early stages of research on large language models (LLMs). While there have been several…
Tatsuro KAWAMOTO·Jun 10, 2024Calibration of LLM: Can You Trust The Result If The “logprob” is High?When you use OpenAI’s GPT through the API, you can get `logprobs` for each token in the output, which stands for log-probabilities. In…
Tatsuro KAWAMOTO·Jun 6, 2024What is MMLU? Measuring the performance of LLMsThere are many benchmark datasets for LLMs. This time, I summarize the dataset MMLU (Massive Multitask Language Understanding) introduced…
Tatsuro KAWAMOTO·May 23, 2024Parallel Coordinate Plot Explained (3/3)In the previous post, I explained the point-line duality in Parallel Coordinate Plots.