"To Mankind Humankind Allkind, and the hope that the war against folly may someday be won, after all." — Isac Asimov Welcome, human or AI. I'm Susana, c...
Behavioral economists assess risk attitudes through choice experiments: subjects choose between certain and probabilistic outcomes (e.g., guaranteed $45 vs. 50% chance of $100). These choices reveal risk aversion, probability weighting and loss aversion. I propose testing whether such methods can measure risk attitudes in LLMs. As proof of concept, I ran a preliminary test with Llama…
Note: This is a quick study, not a full research idea Abstract This study investigates whether backtracking, when an LLM corrects its reasoning path mid-solution, improves performance on mathematical reasoning tasks. Using a subset of 296 problems from GSM8K with the DeepSeek-R1 model, I found backtracking improves success rates by 8.19% overall. This improvement scales with problem difficulty,…
Executive Summary [[Embedding models]] are the foundational layer of modern [[AI]] systems, powering everything from [[search engines]] and content recommendations to language translation and document classification. They're used both as building blocks for [[Large Language Models|LLM]]s and as standalone applications. To understand the scale of embedding model usage: Google processes over 8.5…