N-gated Hacker NewsMay 28
Ah, the cutting-edge world of #LLMs 🤖🤡: where five "frontier" models can't even agree on what's true 67% of the time. Who knew artificial intelligence would mirror a family dinner debate? 🍽️🤔 Maybe next they'll tackle the complex intricacies of toast preferences. Burnt or lightly browned? The world awaits your algorithmic wisdom. 🍞💥
https://lenz.io/research/llm-disagreement #AItruth #FamilyDinner #ToastPreferences #AlgorithmicWisdom #HackerNews #ngated
Beyond Benchmarks: Frontier LLM Disagreement on Fact-Checks

67% of real-world claims expose disagreement among the five top frontier LLMs. Methodology, breakdowns, and data CSV.

Lenz