Capability Evolution
Capability evolution in artificial intelligence focuses on understanding and enhancing the abilities of various AI models, particularly large language models (LLMs), across diverse tasks. Current research emphasizes evaluating these capabilities through novel benchmarks and frameworks, often analyzing model performance under incomplete information or with limited data, and exploring the role of factors like data quality and model architecture (e.g., transformers, state space models). This research is crucial for responsible AI development, informing the creation of more robust and reliable systems with applications ranging from robotics and software engineering to education and scientific research.
Papers
How does Architecture Influence the Base Capabilities of Pre-trained Language Models? A Case Study Based on FFN-Wider and MoE Transformers
Xin Lu, Yanyan Zhao, Bing Qin, Liangyu Huo, Qing Yang, Dongliang Xu
Beyond Specialization: Assessing the Capabilities of MLLMs in Age and Gender Estimation
Maksim Kuprashevich, Grigorii Alekseenko, Irina Tolstykh