CBMM10 Panel: Language and Thought @MITCBMM

Uploaded November 2023 | Updated September 2026, 1 week ago

Is natural language the language of thought? LLMs as models of human language and thought. Are LLMs aligned with neuroscience and with human behavior? What is still missing?

Panel Chair: J. Tenenbaum
Panelists: E. Fedorenko, S. Gershman, P. Isola, E. Spelke, S. Ullman

What to expect when participating in an fMRI study at MIT

CBMM10: CBMM after 10 years — the Quest now

$Model-agnostic Measure of Generalization Difficulty CBMM Research Meeting: Akhilan Boopathy, MIT graduate student in the Fiete Lab Abstract: The measure of a machine learning algorithm is the difficulty of the tasks it can perform, and sufficiently difficult tasks are critical drivers of strong machine learning models. However, quantifying the generalization difficulty of machine learning benchmarks has remained challenging. We propose what is to our knowledge the first model agnostic measure of the inherent generalization difficulty of tasks. Our inductive bias complexity measure quantifies the total information required to generalize well on a task minus the information provided by the data. It does so by measuring the fractional volume occupied by hypotheses that generalize on a task given that they fit the training data. It scales exponentially with the intrinsic dimensionality of the space over which the model must generalize but only polynomially in resolution per dimension, showing that tasks which require generalizing over many dimensions are drastically more difficult than tasks involving more detail in fewer dimensions. Our measure can be applied to compute and compare supervised learning, reinforcement learning and meta-learning generalization difficulties against each other. We show that applied empirically, it formally quantifies intuitively expected trends, e.g. that in terms of required inductive bias, MNIST (less than) CIFAR10 (less than) Imagenet and fully observable Markov decision processes (MDPs) (less than) partially observable MDPs. Further, we show that classification of complex images (less than) few-shot meta-learning with simple images. Our measure provides a quantitative metric to guide the construction of more complex tasks requiring greater inductive bias, and thereby encourages the development of more sophisticated architectures and learning algorithms with more powerful generalization capabilities.$

Functional imaging of the human brain: A window into the organization of the human mind - Part 1

What you see is what IT gets: Responses in primate visual cortex during natural viewing

SGD and Weight Decay Secretly Compress Your Neural Network

Benchmarking Out-of-Distribution Generalization Capabilities of DNN-based Encoding Models for the...

Panel Discussion: Open Problems in the Theory of Deep Learning

MITCBMM |

CBMM10 Panel: Language and Thought