Julius Cheng

PhD at @cambridgenlp, former staff ML eng @otter_ai, MSCS @Stanford, linguistics & cogsci @UCBerkeley

[Vilém Zouhar @ EMNLP @zouharvi]

There are many things machine translation still can't do. Help us steer the next direction by contributing hard-to-translate inputs (and be on a cool paper).

[Vilém Zouhar @ EMNLP @zouharvi]

Quality estimation (automated metrics) are amazing. Truly. We would like to use them everywhere. That gets compute-expensive very quickly. We also don't know when they don't know.

In "Early-Exit and Instant Confidence Translation Quality Estimation" (at EACL26) we fix that.

[Bryan Eikema @BryanEikema]

The Second Workshop on Uncertainty-Aware NLP will be co-located with #EMNLP2025 this November in Suzhou, China! We’re excited for another great gathering of uncertainty enthusiasts in NLP. For more details, please visit our website: https://t.co/3O1DUps5YC

[Julius Cheng @julius_gulius]

@glampouras_NLP Congrats to you and the team. These papers look really cool!

[Vilém Zouhar @ EMNLP @zouharvi]

The paper comes with mysterious fig 1. 👁️

[Julius Cheng @julius_gulius]

Our experiments are on machine translation, but this method works with any generator + reranker setup!

Eager to get feedback, and happy reranking!

[Julius Cheng @julius_gulius]

Happy to announce that our work "A Bayesian Optimization Approach to Machine Translation" was accepted to NAACL 2025!

Special thanks to @ufal_cuni for organizing MT Marathon 2025 where I was able to team up with @MaikeZufle @zouharvi!

Explainer below:

https://t.co/KsQXX7kwUI

[Julius Cheng @julius_gulius]

Bonus: we show how to use multi-fidelity Bayesian optimization to use a smaller and faster proxy scoring model to make even better decisions. We get the best performance by training a distilled model from our main CometKiwi model.

[Institute of Formal and Applied Linguistics @ufal_cuni]

@OndrejBojar @prajdabre1 @zouharvi @esalesk We continued with great talks by @RicardoRei7, @julius_gulius, @onadegibert, @josephnlp, @very_laurie, @khetamalsharou, and @Lam19Tk. Thanks for coming to Prague 🙌. ➕ Throughout the week, people worked on 👉7 🤩 exciting hacking projects.

[Julius Cheng @julius_gulius]

Massive apologies for the delay on this, but I've finally finished open sourcing the code for our EMNLP 2023 Best Short Paper winning paper on efficient MBR. This codebase also provides a solid foundation if you're experimenting with sampling-based MBR!

https://t.co/5LQdrPyMTi

[Julius Cheng @julius_gulius]

@tozefarinhas It was great to meet you at EMNLP!

[Julius Cheng @julius_gulius]

When I was young, they told me "Julius you will never get Best Short Paper at EMNLP"

Today I proved them wrong

[EMNLP 2026 @emnlpmeeting]

EMNLP 2023 Best Short Paper

Faster Minimum Bayes Risk Decoding with Confidence-based Pruning

(Julius Cheng, Andreas Vlachos)

https://t.co/FX4udDiaL0

#EMNLP2023 #NLProc

[Julius Cheng @julius_gulius]

@wilkeraziz Thanks! I credit you and Bryan for getting me interested in MBR!

[Julius Cheng @julius_gulius]

@BryanEikema @vlachos_nlp You're no stranger to MBR papers getting awards! Thanks for writing the paper that got me interested in MBR!