Research Warnings on the Limits of AI Language Models: Over 8K Context Performance Halves, Conceptual Reasoning Becomes a Challenge

A recent study released by the University of Munich, the Munich Machine Learning Center, and Adobe Research shows that 12 top AI language models, including GPT-4o, Gemini1.5Pro, and Llama-3.3-70B, face significant performance degradation in long-text conceptual reasoning tasks. Despite these models supporting context processing of at least 128,000 tokens, their ability to establish deep logical connections still has fundamental limitations. The research team developed NOLIMA (No Text Matching).