Morris Alper

I am a PhD student in the Department of Computer Science at Cornell University (Cornell Tech campus), supervised by Dr. Hadar Averbuch-Elor. I research multimodal learning, combining vision with modalities such as language and three-dimensional data for complex tasks requiring high-level semantic understanding.

I received my MSc with honors from Tel Aviv University in Computer Science and my BSc from MIT with a dual major in mathematics and linguistics. I was a research scientist intern at Meta in 2024, and have industry experience working as a data scientist and teaching machine learning to industry professionals.

Email  /  CV  /  Scholar  /  Twitter  /  Bluesky  /  GitHub  /  LinkedIn  /  ORCID

We introduce WAFFLE, a novel multimodal floorplan understanding dataset of nearly 20K diverse floorplan images and metadata, and we show that this data enables progress on new building understanding tasks.

We propose a framework for addressing open-vocabulary hallucinations in image captioning models, including a new benchmark and a reinforcement learning-based method to reduce such hallucinations.

We show that foundation VLMs like CLIP model visual-semantic hierarchies, proposing the Radial Embedding framework for probing and optimizing this knowledge and the HierarCaps dataset of ground-truth image caption hierarchies.

We quantify image caption concreteness using information loss in foundation vision-language models, and use this score to filter web-scale multimodal datasets.

We learn a semantic localization field for textual descriptions over collections of in-the-wild images depicting a large-scale scene.

By generating images using prompts containing pseudowords (nonsense words) and analyzing their shapes, we show that AI image generation models show sound-shape associations similar to those known from human psychology.

We model human-human interaction understanding in images as free text generation, provide a new benchmark and show how to learn this with weak supervision from Internet image captions.

We find that multimodally trained text encoders outperform unimodally trained text encoders on visual reasoning in text.

• Our work on the Kiki-Bouba effect and sound symbolism in AI models was featured in the Haaretz newspaper (Hebrew language). I was also interviewed about this work on the Kan Tarbut morning radio program (recording link, 46:30-60:00, Hebrew language).

• UNIKUD: Adding Vowels to Hebrew Text with Deep Learning
• TaatikNet: Sequence-to-Sequence Learning for Hebrew Transliteration

• As a high school student, I qualified to join the USA team at the International Linguistics Olympiad (IOL) for three years (2008, 2009, 2011). In particular, in 2011 I won 1st place (gold medal) on individual and team rounds in 2011 as well as receiving the Zhurinski Memorial Prize and a best solution award. I also received a silver medal (tied for 6th place) in 2008 and our team won the team round in 2009.

Page format adapted from Jon Barron's template.