The field of artificial intelligence is witnessing significant developments in Class Incremental Learning (CIL) and Large Language Models (LLMs). Researchers are exploring innovative methods to improve the performance and reliability of CIL models, such as the use of Increment Vector Transformation (IVT) to mitigate catastrophic forgetting. Meanwhile, LLMs are being increasingly applied in various domains, including education and social science, with a growing focus on evaluating their risks and benefits. The social science of LLMs is emerging as a distinct area of research, examining the interactions between LLMs and human societies. Noteworthy papers in this area include 'Closing the Oracle Gap: Increment Vector Transformation for Class Incremental Learning', which proposes a novel framework for CIL, and 'The Social Laboratory: A Psychometric Framework for Multi-Agent LLM Evaluation', which introduces a new evaluation framework for LLMs in interactive environments. Overall, the field is moving towards more robust and reliable AI systems, with a growing emphasis on understanding their social implications.
Advancements in Class Incremental Learning and Large Language Models
Sources
Closing the Oracle Gap: Increment Vector Transformation for Class Incremental Learning
From Superficial Outputs to Superficial Learning: Risks of Large Language Models in Education
The Lie of the Average: How Class Incremental Learning Evaluation Deceives You?
The Emergence of Social Science of Large Language Models
A Cartography of Open Collaboration in Open Source AI: Mapping Practices, Motivations, and Governance in 14 Open Large Language Model Projects
The Flaw of Averages: Quantifying Uniformity of Performance on Benchmarks
Evaluating the Use of Large Language Models as Synthetic Social Agents in Social Science Research
The Hunger Game Debate: On the Emergence of Over-Competition in Multi-Agent Systems
RELATE-Sim: Leveraging Turning Point Theory and LLM Agents to Predict and Understand Long-Term Relationship Dynamics through Interactive Narrative Simulations
Emergent evaluation hubs in a decentralizing large language model ecosystem
The Social Laboratory: A Psychometric Framework for Multi-Agent LLM Evaluation
Position: Privacy Is Not Just Memorization!
To Mask or to Mirror: Human-AI Alignment in Collective Reasoning