Cheng Qian

Cheng Qian, Xinran Zhao, Sherry Tongshuang Wu · tsinghua

Large language models (LLMs) acquire extensive knowledge during pre-training, known as their parametric knowledge. However, in order to remain up-to-date and align with human instructions, LLMs inevitably require external knowledge during their interactions with users. This raises a crucial question: How will LLMs respond when external knowledge interferes with their parametric knowledge? To investigate this question, we propose a framework that systematically elicits LLM parametric knowledge and introduces external knowledge. Specifically, we uncover the impacts by constructing a parametric knowledge graph to reveal the different knowledge structures of LLMs, and introduce external knowledge through distractors of varying degrees, methods, positions, and formats. Our experiments on both black-box and open-source models demonstrate that LLMs tend to produce responses that deviate from their parametric knowledge, particularly when they encounter direct conflicts or confounding changes of information within detailed contexts. We also find that while LLMs are sensitive to the veracity of external knowledge, they can still be distracted by unrelated information. These findings highlight the risk of hallucination when integrating external knowledge, even indirectly, during interactions with current LLMs. All the data and results are publicly available.

Cheng Qian

Understanding the basic operational logics of the nervous system is essential to advancing neuroscientific research. However, theoretical efforts to tackle this fundamental problem are lacking, despite the abundant empirical data about the brain that has been collected in the past few decades. To address this shortcoming, this document introduces a hypothetical framework for the functional nature of primitive neural networks. It analyzes the idea that the activity of neurons and synapses can symbolically reenact the dynamic changes in the world and thus enable an adaptive system of behavior. More significantly, the network achieves this without participating in an algorithmic structure. When a neuron's activation represents some symbolic element in the environment, each of its synapses can indicate a potential change to the element and its future state. The efficacy of a synaptic connection further specifies the element's particular probability for, or contribution to, such a change. As it fires, a neuron's activation is transformed to its postsynaptic targets, resulting in a chronological shift of the represented elements. As the inherent function of summation in a neuron integrates the various presynaptic contributions, the neural network mimics the collective causal relationship of events in the observed environment.

Kehan Long, Cheng Qian, Jorge Cortés et al.

Control barrier functions are widely used to enforce safety properties in robot motion planning and control. However, the problem of constructing barrier functions online and synthesizing safe controllers that can deal with the associated uncertainty has received little attention. This paper investigates safe navigation in unknown environments, using onboard range sensing to construct control barrier functions online. To represent different objects in the environment, we use the distance measurements to train neural network approximations of the signed distance functions incrementally with replay memory. This allows us to formulate a novel robust control barrier safety constraint which takes into account the error in the estimated distance fields and its gradient. Our formulation leads to a second-order cone program, enabling safe and stable control synthesis in a priori unknown environments.