8 Cool New AI Papers: From Anxious LLMs to Fake News Detection
- By Bilal Hafeez
This article is only available to Macro Hive subscribers. Sign-up to receive world-class macro analysis with a daily curated newsletter, podcast, original content from award-winning researchers, cross market strategy, equity insights, trade ideas, crypto flow frameworks, academic paper summaries, explanation and analysis of market-moving events, community investor chat room, and more.
- Assessing and Alleviating State Anxiety in LLMs: Finds that trauma prompts can change the ‘state’ anxiety of LLMs, while relaxation ones reduce anxiety!
- Structured Outputs Enable General-Purpose LLMs to Be Medical Experts: The paper does away with fine-tuning an LLM on medical data and instead introduces a seven-step cognitive process (chain-of-thought) to answer questions. E.g., ‘understand question’, ‘recall relevant medical knowledge’, etc. Performs well on benchmarks. See Figure 1 below.
- Expert Prompting: Instructing Large Language Models to Be Distinguished Experts: Clever use of in-context learning to generate automated expert prompts (rather than having to describe the expert yourself). Leads to much better answers to questions.
- Agentic AI Needs a Systems Theory: This thought-provoking paper argues for a systems approach to understanding agent behaviour rather than individual agent behaviours.
- Unseen Fake News Detection Through Casual Debiasing: I like the use of causal reasoning to get rid of fake news and bias from training sets. Could be used for finding errors more generally in training sets.
- ToolFuzz – Automated Agent Tool Testing: We are working on agentic LLMs which requires good documentation on the codebase of the tools. This ToolFuzz approach looks promising for testing for documentation errors.
- How Diversely Can Language Models Solve Problems? Exploring the Algorithmic Diversity of Model-Generated Code: Finds that LLM models have little diversity in solving problems compared to human solutions. They find that raising the temperature of models beyond 1.0 could increase diversity.
- ChroKnowledge: Unveiling Chronological Knowledge of Language Models in Multiple Domains: Great attempt at assessing how the chronological knowledge of LLMs evolves. Most LLM models struggle with capturing time.
(The commentary contained in the above article does not constitute an offer or a solicitation, or a recommendation to implement or liquidate an investment or to carry out any other transaction. It should not be used as a basis for any investment decision or other decision. Any investment decision should be based on appropriate professional advice specific to your needs. You are not permitted to publish, transmit, or otherwise reproduce this information, in whole or in part, in any format to any third party without the express written consent of Macro Hive. This includes providing or reproducing this information, in whole or in part, as a prompt.)