Rohit Saxena

rohit.saxena@ed.ac.uk

University of Edinburgh

I am a final year PhD student at the University of Edinburgh. I am advised by Prof Frank Keller. I also collaborate with Prof Pasquale Minervini and Prof Hao Tang. I am affiliated with EdinburghNLP and CDT NLP.

My research interests broadly lie in natural language understanding and multimodal learning, with a focus on long-context modeling and the intersection of vision and language.

Previously, I worked as a Researcher at Tata Research Development and Design Center (TRDDC), TCS Research, Pune India. At TRDDC, I was part of Media and Entertainment Research group and was advised by Niranjan Pedanekar. My work there involved emotion detection in dialogues, analysing TV viewers’ attention, and style transfer for advertisements. Additionally, I interned as an Applied Scientist at Amazon Web Services (AWS) in Seattle, where I worked on mitigating hallucinations in large language models.

In my free time, I enjoy photography.

news

Jun 01, 2025	Paper accepted at TACL 2025.
Mar 13, 2025	Paper accepted at ICLR 2025 Workshop.
Jan 22, 2025	Paper accepted at NAACL 2025.
Aug 14, 2024	Presented work at ACL 2024.
Jun 19, 2024	Presented work at NAACL 2024

selected publications

TACL 2025

End-to-End Long Document Summarization using Gradient Caching

Rohit Saxena, Hao Tang, and Frank Keller

2025
ICLR 2025 Workshop

Lost in Time: Clock and Calendar Understanding Challenges in Multimodal LLMs

Rohit Saxena, Aryo Pradipta Gema, and Pasquale Minervini

2025
Under Review

PosterSum: A Multimodal Benchmark for Scientific Poster Summarization

Rohit Saxena, Pasquale Minervini, and Frank Keller

2025
ACL 2024

MovieSum: An Abstractive Summarization Dataset for Movie Screenplays

Rohit Saxena, and Frank Keller

In Findings of the Association for Computational Linguistics ACL 2024, Aug 2024

Abs

Movie screenplay summarization is challenging, as it requires an understanding of long input contexts and various elements unique to movies. Large language models have shown significant advancements in document summarization, but they often struggle with processing long input contexts. Furthermore, while television transcripts have received attention in recent studies, movie screenplay summarization remains underexplored. To stimulate research in this area, we present a new dataset, MovieSum, for abstractive summarization of movie screenplays. This dataset comprises 2200 movie screenplays accompanied by their Wikipedia plot summaries. We manually formatted the movie screenplays to represent their structural elements. Compared to existing datasets, MovieSum possesses several distinctive features: 1) It includes movie screenplays which are longer than scripts of TV episodes. 2) It is twice the size of previous movie screenplay datasets. 3) It provides metadata with IMDb IDs to facilitate access to additional external knowledge. We also show the results of recently released large language models applied to summarization on our dataset to provide a detailed baseline.