Large language models are successful in answering factoid questions but are also prone to hallucination.We investigate the phenomenon of LLMs possessing correct answer knowledge yet still hallucinating from the perspective of inference dynamics, an area not previously covered in studies on hallucinations.We are able to conduct this analysis via two key ideas.First, we identify the factual questions that query the same triplet knowledge but result in different answers. The difference between the model behaviors on the correct and incorrect outputs hence suggests the patterns when hallucinations happen.Second, to measure the pattern, we utilize mappings from the residual streams to vocabulary space.We reveal the different dynamics of the output token probabilities along the depths of layers between the correct and hallucinated cases. In hallucinated cases, the output token’s information rarely demonstrates abrupt increases and consistent superiority in the later stages of the model.Leveraging the dynamic curve as a feature, we build a classifier capable of accurately detecting hallucinatory predictions with an 88% success rate. Our study shed light on understanding the reasons for LLMs’ hallucinations on their known facts, and more importantly, on accurately predicting when they are hallucinating.

The multi-modality nature of human communication has been utilized to enhance the performance of language modeling-related tasks. Driven by the development of large-scale end-to-end learning techniques and the availability of multi-modal data, it becomes possible to represent non-verbal communication behaviors through joint-learning, and directly study their interaction with verbal communication. However, there is still gaps in existing studies to better address the underlying mechanism of how non-verbal expression contributes to the overall communication purpose. Therefore, we explore two questions using mixed-modal language models trained against monologue video data: first, whether incorporating gesture representations can improve the language model’s performance (perplexity); second, whether spontaneous gestures demonstrate entropy rate constancy (ERC), which is an empirical pattern found in most verbal language data that supports the rational communication assumption from Information Theory. We have positive and interesting findings for both questions: speakers indeed use spontaneous gestures to convey “meaningful” information that enhances verbal communication, which can be captured with a simple spatial encoding scheme. More importantly, gestures are produced and organized rationally in a similar way as words, which optimizes the communication efficiency.

Verbal communication is companied by rich non-verbal signals. The usage of gestures, poses, and facial expressions facilitates the information transmission in verbal channel. However, few computational studies have explored the non-verbal channels with finer theoretical lens. We extract gesture representations from monologue video data and train neural sequential models, in order to study the degree to which non-verbal signals can effectively transmit information. We focus on examining whether the gestures demonstrate the similar pattern of entropy rate constancy (ERC) found in words, as predicted by Information Theory. Positive results are shown to support the assumption, which leads to the conclusion that speakers indeed use simple gestures to convey information that enhances verbal communication, and the production of non-verbal information is rationally organized.