Muhammad Cendekia Airlangga

Hello, I’m Cendekia! 👋

I am a first-year PhD student in the Department of Natural Language Processing at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), supervised by Prof. Kentaro Inui. My research sits at the intersection of causality and interpretability, with a central question:

How can causality help us understand what language models compute, and how can we make language models better at understanding cause and effect?

I am particularly interested in:

Interpreting models with causality: Using causal tools such as interventions, causal mediation and causal abstraction to find which internal mechanisms actually drive a model’s behaviour.
Causal understanding in language models: Why do language models struggle with causal reasoning, and how can insights from interpretability help them reason better about cause and effect?
Low-resource languages: How can we build robust language technology for Indonesian languages such as Sundanese and Javanese?

I received my MSc in Natural Language Processing from MBZUAI and my Bachelor of Engineering in Electrical Engineering (cum laude) from Institut Teknologi Sepuluh Nopember (ITS), Surabaya, Indonesia 🇮🇩. I was also a research intern in the Natural Language Understanding Team at RIKEN AIP in Japan 🇯🇵.

News

Oct 16, 2026	Our paper “Sparse Long Memory and Transition Bottlenecks Shape Positional Recall in Mamba Language Models” has been accepted to Findings of EMNLP 2026! See you in Budapest 🇭🇺
Jun 04, 2026	Our paper “A Unified Safety Subspace Exists in Speech Language Models” has been accepted at Interspeech 2026! 🎉
May 07, 2026	Graduated with an MSc in Natural Language Processing from MBZUAI.
Apr 07, 2026	Successfully defended my MSc thesis on vision-language model interpretability. 🎓
Nov 20, 2025	Gave a guest lecture on the interpretability of large language models for IF5153 Natural Language Processing at Institut Teknologi Bandung (ITB).
Nov 08, 2025	Our paper “ASR Under Noise: Exploring Robustness for Sundanese and Javanese” received the Best Paper Award at the 9th WiNLP Workshop at EMNLP 2025! 🏆
Oct 02, 2025	Presented our SSM interpretability project, “Emergence of Primacy and Recency Effect in Mamba”, at the RIKEN AIP LLM×ML Workshop in Japan. 🇯🇵

Latest Posts

Oct 15, 2026	Hello, world

Selected Publications

Interspeech

A Unified Safety Subspace Exists in Speech Language Models

Nurdaulet Mukhituly, Muhammad Cendekia Airlangga, Rifo Ahmad Genadi, Nhi Hoai Doan, Amirbek Djanibekov, Samuel Munachiso Nwadike, Zangir Iklassov, and Kentaro Inui

In Interspeech 2026, 2026

Bib PDF

@inproceedings{airlangga2026safety,
  title = {A Unified Safety Subspace Exists in Speech Language Models},
  author = {Mukhituly, Nurdaulet and Airlangga, Muhammad Cendekia and Genadi, Rifo Ahmad and Doan, Nhi Hoai and Djanibekov, Amirbek and Nwadike, Samuel Munachiso and Iklassov, Zangir and Inui, Kentaro},
  booktitle = {Interspeech 2026},
  year = {2026},
}

WiNLP

ASR Under Noise: Exploring Robustness for Sundanese and Javanese

Muhammad Cendekia Airlangga, Salsabila Zahirah Pranida, Rifo Ahmad Genadi, and Shady Shehata

In Proceedings of the 9th Widening NLP Workshop (WiNLP) at EMNLP 2025, 2025

Best Paper arXiv Bib HTML

Best Paper Award

@inproceedings{airlangga2025asr,
  title = {ASR Under Noise: Exploring Robustness for Sundanese and Javanese},
  author = {Airlangga, Muhammad Cendekia and Pranida, Salsabila Zahirah and Genadi, Rifo Ahmad and Shehata, Shady},
  booktitle = {Proceedings of the 9th Widening NLP Workshop (WiNLP) at EMNLP 2025},
  year = {2025},
  url = {https://aclanthology.org/2025.winlp-main.16/},
}

EMNLP

Sparse Long Memory and Transition Bottlenecks Shape Positional Recall in Mamba Language Models

Muhammad Cendekia Airlangga, Hilal AlQuabeh, Munachiso S. Nwadike, and Kentaro Inui

In Findings of the Association for Computational Linguistics: EMNLP 2026, Budapest, Hungary 🇭🇺, 2026

arXiv Bib PDF

@inproceedings{airlangga2026mamba,
  title = {Sparse Long Memory and Transition Bottlenecks Shape Positional Recall in Mamba Language Models},
  author = {Airlangga, Muhammad Cendekia and AlQuabeh, Hilal and Nwadike, Munachiso S. and Inui, Kentaro},
  booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2026},
  year = {2026},
  location = {Budapest, Hungary 🇭🇺},
  eprint = {2506.15156},
  archiveprefix = {arXiv},
}