Topic: proactive

A new study warns that fine-tuning large language model (LLM)-based proactive dialogue systems with external data can introduce backdoor vulnerabilities, which existing research has largely overlooked. This matters because such systems are used for goal-directed support and could be exploited if attackers sneak in malicious instructions through the training data.

arxiv.org↗

A new approach in LLM agent memory systems focuses on prospective memory, enabling agents to retain future-oriented intentions and execute them at the right time, moving beyond just recalling past information. This development is crucial as it enhances the functionality of LLMs, making them more adaptable and effective in various applications that require forward-planning capabilities.

arxiv.org↗