11 papers
SCITUNE: Aligning Large Language Models with Human-Curated Scientific Multimodal Instructions
Sameera Horawalavithana, Sai Munikoti, Ian Stewart +2
Instruction finetuning is a popular paradigm to align large language models (LLM) with human intent. Despite its popularity, this idea is less explored in improving LLMs to align e…
MINAR: Mechanistic Interpretability for Neural Algorithmic Reasoning
Jesse He, Helen Jenne, Max Vargas +4
The recent field of neural algorithmic reasoning (NAR) studies the ability of graph neural networks (GNNs) to emulate classical algorithms like Bellman-Ford, a phenomenon known as…
Can Neural Networks Learn Small Algebraic Worlds? An Investigation Into the Group-theoretic Structures Learned By Narrow Models Trained To Predict Group Operations
Henry Kvinge, Andrew Aguilar, Nayda Farnsworth +4
While a real-world research program in mathematics may be guided by a motivating question, the process of mathematical discovery is typically open-ended. Ideally, exploration neede…
Saddle-Free Guidance: Improved On-Manifold Sampling without Labels or Additional Training
Eric Yeats, Darryl Hannan, Wilson Fearn +3
Score-based generative models require guidance in order to generate plausible, on-manifold samples. The most popular guidance method, Classifier-Free Guidance (CFG), is only applic…
Even with AI, Bijection Discovery is Still Hard: The Opportunities and Challenges of OpenEvolve for Novel Bijection Construction
Davis Brown, Jesse He, Helen Jenne +2
Evolutionary program synthesis systems such as AlphaEvolve, OpenEvolve, and ShinkaEvolve offer a new approach to AI-assisted mathematical discovery. These systems utilize teams of…
A Connection Between Score Matching and Local Intrinsic Dimension
Eric Yeats, Aaron Jacobson, Darryl Hannan +4
The local intrinsic dimension (LID) of data is a fundamental quantity in signal processing and learning theory, but quantifying the LID of high-dimensional, complex data has been a…