top of page

Challenge the Frontier

The Theoretical Research Team is your chance to conduct term-time machine learning research with your friends! HUMIC sponsors innovative research ideas with all the resources, funding, clout, support network, and insider knowledge that you need. Every semester, the team aims to submit a high quality paper to top machine learning conferences such as NeurIPS, ICML, and AAAI. The goal is to contribute meaningfully to how we think about machine learning, and the research generally focuses on developing nascent theoretical explanations and supporting them with rigorous empirical results.

In addition, we connect student researchers to labs at Harvard looking for undergraduate helpers. If you would like access to our network and resources page, join our email list!

If interested, please reach out to us at

Past Projects

Project Team: Leonard Tang, Alexander Cai, Steve Li, Jason Wang

Jokes are intentionally written to be funny, but not all jokes are created the same. Some jokes may be fit for a classroom of kindergarteners, but others are best reserved for a more mature audience. While recent work has shown impressive results on humor detection in text, here we instead investigate the more nuanced task of detecting humor subtypes, especially of the less innocent variety. To that end, we introduce a novel jokes dataset filtered from Reddit and solve the subtype classification task using a finetuned Transformer dubbed the Naughtyformer. Moreover, we show that our model is significantly better at detecting offensiveness in jokes compared to state-of-the-art methods.

Screenshot 2023-04-18 at 8.06.54 PM.png

Raven's Progressive Matrices

Project Team: Deepak Singh, Chloe Loughridge, Darius Lam

Raven's Progressive Matrices (RPM) is a commonly-used intelligence test for abstract visual reasoning skills. While humans are quite good at it, computers are notoriously lacking in their performance on this test. Students at HUMIC Labs developed novel deep learning methods to tackle this challenge.

bottom of page