paper

Modeling Complex Behaviors: Multi-Personality Composition and Dynamic Switching in Vision-Language Models

paperactiveprovisionalmodeling-complex-behaviors-multi-personality-composition-and-dynamic-switching-in-vision-language-models-99128d23·1 events·first seen 7d ago

Aliases: Modeling Complex Behaviors: Multi-Personality Composition and Dynamic Switching in Vision-Language Models

More like this (12)

Vision-Language Models Vision-Language-Action models visual language model TempoVLA: Learning Speed-Controllable Vision-Language-Action Policies Vision-Language-Action model LabVLA: Grounding Vision-Language-Action Models in Scientific Laboratories Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models multi-turn language models Unified Multimodal Models (UMMs)Language Modeling Loss Multimodal Large Language Models Exploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Models

Recent events (1)

4arXiv · cs.CL·7d ago·source ↗

Systematic evaluation of multi-personality conditioning and dynamic switching in vision-language models

This paper introduces explicit personality conditioning for multimodal large language models (MLLMs) and proposes an evaluation framework covering single-personality induction, multi-personality composition, and dynamic personality switching. Experiments reveal that personality induction improves image captioning but degrades performance on precise reasoning tasks like VQA. The authors find balancing and residual effects during multi-trait composition and switching, and show that existing prompt-based personality induction methods transfer poorly to multimodal settings.

Evaluation and Benchmarking Multimodal Progress Modeling Complex Behaviors: Multi-Personality Composition and Dynamic Switching in Vision-Language Models