Close
Experimental Psychology Seminar - Refael Tikochinski
03 December 2024, 1:00 pm–2:00 pm
Incremental Accumulation of Linguistic Context in Artificial and Biological Neural Networks
Event Information
Open to
- All
Organiser
-
Antonietta Esposito
Location
-
30526 Bedford WayLondonWC1H 0AP
Abstract: Accumulated evidence suggests that Large Language Models (LLMs) are beneficial in predicting neural signals related to narrative processing. The way LLMs integrate context over large timescales, however, is fundamentally different from the way the brain does it. In my study, we show that unlike LLMs that apply parallel processing of large contextual windows, the incoming context to the brain is limited to short windows of a few tens of words. We hypothesize that whereas lower-level brain areas process short contextual windows, higher-order areas in the default-mode network (DMN) engage in an online incremental mechanism where the incoming short context is summarized and integrated with information accumulated across long timescales. Consequently, we introduce a novel LLM that instead of processing the entire context at once, it incrementally generates a concise summary of previous information. As predicted, we found that neural activities at the DMN were better predicted by the incremental model, and conversely, lower-level areas were better predicted with short-context-window LLM.
Zoom Link: https://ucl.zoom.us/j/97450915187?pwd=tPaFVJot6eusIQobxw1BHa9waU2TcD.1
Meeting ID: 974 5091 5187
Passcode: 418905
About the Speaker
Refael Tikochinski
at Experimental Psychology, UCL
I am a cognitive neuroscientist and a fellow of the Israel Council for Higher Education. I am interested in the way semantic information is organized and represented in the human brain, and how this organization helps individuals build their own mental models of the world. Specifically, I investigate how exposure to new knowledge influences internal mental representations and the subsequent effects on future information-seeking behaviors. I use advanced computational research methods, primarily from the field of Natural Language Processing (NLP), alongside more “classical” behavioral and neuroimaging methods from the field of psychology.
I completed my PhD at the Technion – Israel Institute of Technology in Haifa, Israel, under the joint supervision of Prof. Roi Reichart (Faculty of Data and Decision Sciences, Technion) and Prof. Uri Hasson (Princeton Neuroscience Institute, Princeton University). My PhD dissertation focused on various uses of large language models (LLMs) to investigate how the human brain incorporates context and perspectives into language processing. Prior to that, I completed my Bachelor’s and Master’s degrees in Psychology at the Hebrew University of Jerusalem, Israel.