Contributed Talks: “Neural mechanisms of learning and memory: from synapses to systems” - CCN 2025

Value of the Information & Strength of the Argument

The value of the information is high, as the talks present novel computational models and experimental paradigms that advance understanding of learning and memory. The argumentation is solid, with each speaker providing clear hypotheses, methods, and results. For instance, Huang Ham’s talk uses a computational model to benchmark optimal strategies and compares them with human data, showing adaptive division of labor. Zoran Tiganj’s model offers a novel perspective on synaptic weight representation, with potential implications for episodic memory. Kate Nussenbaum’s work systematically characterizes counterfactual learning, providing a new task and model to measure individual differences. The presentations are well-structured and supported by data, though some talks are brief and lack extensive detail.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is generally high, with presentations based on peer-reviewed research or work in progress at a reputable conference. The speakers cite relevant literature and provide links to abstracts and posters. The title accurately reflects the content, as it is a session on neural mechanisms of learning and memory. The sources cited are primarily the speakers’ own work and conference materials, which are appropriate. The video does not include external references beyond the conference context, but the presentations themselves are grounded in established research. The adequacy between title and content is strong.

The title accurately reflects the content: a session of contributed talks on neural mechanisms of learning and memory, covering topics from synaptic to systems level.

8/10

The video presents peer-reviewed research from a reputable conference (CCN 2025), with clear methodology and results. The talks are given by researchers from recognized institutions. However, the video is a recording of a session, and the content is not independently verified here.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction to the session and first talk by Huang Ham on collaborative encoding of visual working memory.
• 05:00Huang Ham explains the spatial working memory task and the computational model for optimal division of labor.
• 10:00Huang Ham presents results showing adaptive division of labor, even when benefit is minimal.
• 15:00Zoran Tiganj begins his talk on computational model for episodic timelines based on synaptic decay rates.
• 20:00Tiganj explains the vector representation of synaptic weights and the timeline encoding.
• 25:00Tiganj illustrates the model with a video example and discusses implications for episodic memory.
• 30:00Kate Nussenbaum presents her work on strategic counterfactual learning.
• 35:00Nussenbaum describes the battle cards game and the model fitting results showing upward counterfactual learning.
• 40:00John C Bowler presents on abstraction learning in recurrent networks.
• 45:00Mingze Li Leukos presents on reward-prediction-error-guided attention and learning curves.

• Abstract for Huang Ham's talk — Referenced in the talk as a link to the abstract.
• Poster C70 — Mentioned by Huang Ham for bonus results.
• Poster C73 — Mentioned by Zoran Tiganj for more details.

• CCN 2025 Conference Website — Official conference page confirming the session and speakers.