ZurichNLP #13

Besta Maciej (ETH Zurich) on "Effective and Efficient LLM Ecosystems with Graph of Thoughts and Beyond": Graph of Thoughts (GoT) is a framework that advances reasoning and prompting capabilities in large language models (LLMs) beyond those offered by paradigms such as Chain-of-Thought or Tree of Thoughts (ToT). The key idea and primary advantage of GoT is the ability to model the information generated by an LLM as an arbitrary graph, where units of information ("LLM thoughts") are vertices, and edges correspond to dependencies between these vertices. This approach enables combining arbitrary LLM thoughts and whole reasoning chains into synergistic outcomes, preserving the experience from failures instead of discarding it, distilling the essence of whole networks of thoughts, or enhancing thoughts using feedback loops. We illustrate that GoT offers advantages in the outcome accuracy and token cost over state of the art on different tasks such as keyword counting, NDA merging, or sorting.…