Philip John Gorinski

PhD, Natural Language Processing. Research Scientist Huawei. Geek and nerd, coffee-driven sportsman. Party goer. Sometimes cook.

[Gerasimos Lampouras @glampouras_NLP]

Looking to grow our NLP team in London - Huawei Noah's Ark Lab! We have positions for Research Scientists (permanents) and Engineers (contractors), where they will conduct academic and applied research in NLP and ML. Details below:

https://t.co/dDxE97BgRl
https://t.co/LGN9snl3yb

[Philip John Gorinski @xylankant]

Check out what my colleagues' @fenchr
, Guchun Zhang, and @glampouras_NLP have just dropped! Modality-relative Pre-training, certainly one of the highlights of this year's EACL 2024!
#Huawei #NLP #LLM #eacl24 #persevere

[Fenia Christopoulou @fenchri]

Happy to share our work on "Text2Code Generation with Modality-relative Pre-training" w/ @gcsanity & @glampouras_NLP, accepted at #EACL2024! 🎉

We propose to treat code and natural language as
different modalities.

📜https://t.co/fOYsUARpve
💻coming soon (pending int. review)

[Gerasimos Lampouras @glampouras_NLP]

Very happy to be co-organizing this workshop and shared task with the fine folks at @TCD and @AdaptCentre! Please help us spread the word of our research track and shared task:

Paper deadline: Dec 18th
ARR commitment deadline: Jan 17th
Shard task deadline: Jan 20th

[Philip John Gorinski @xylankant]

Preprint available at https://t.co/ZrtMkWHnYg ; code and model will be made available.

#Machinelearning #AI #ML #ReinforcementLearning #LLM #PLM #CodeSyntheis #Huawei

[Philip John Gorinski @xylankant]

Proud to announce our paper on "Automatic Unit Test Data Generation and Actor-Critic Reinforcement Learning for Code Synthesis" has been accepted to Findings of #EMNLP2023 .
This is joint work with Matthieu Zimmer, @glampouras_NLP , Derrick Goh Xin Deik, and @iiacobacNLP .

[Philip John Gorinski @xylankant]

Our approach makes use of Test annotated data to tune a function-level Code Synthesis LM.
Crucially, we find keeping the Critic in sync with the Policy yields better results than pretraining and freezing the Critic.
Use of our augmentation data further improves model performance.

[Philip John Gorinski @xylankant]

Take two sets P and N of strings, a cost function, and go find a minimal regex that covers all of P but rejects all of N.

Turns out, this is not as easy as it sounds for modern Deep Learning-based approaches like Large (code) Language Models!

#persevere #research

[Philip John Gorinski @xylankant]

Check out the Regular Expression Inference Challenge, organised by yours truly with Mojtaba Valizadeh, @iiacobacNLP , and Martin Berger: https://t.co/SVvAZd90Um
https://t.co/64APsqHWt1

#machinelearning #inference #challenge #AI #ML #llm #llms #huawei

[Philip John Gorinski @xylankant]

https://t.co/BapzhzR1zh

Google presents the "Unlearning", or, "Oh Snap We Trained on Massive Private Data and now People are Angry"-Challenge.

Still, pretty cool task :)

#nlp #ai #machinelearning etc.

[Iacobacci et al. @iiacobacNLP]

For Interns this is the link

https://t.co/Pf5SfvsXn3

[Iacobacci et al. @iiacobacNLP]

Our wonderful NLP team at Huawei Noah’s Ark Lab is looking for Research Engineers and Research interns.

We are looking for people to work on LLMs for general code and reasoning-related tasks.

Please apply here

https://t.co/MpdYTPEX2O

[Philip John Gorinski @xylankant]

so, I guess I'm on #Mastodon now too, 'cause looks like #mastodonmigration is en-vogue now.
you can follow @[email protected] if you like... expect even fewer toots than tweets though, I reckon... or expect tweets to go down and toots up... will I cross-post? who knows. okbye

[Philip John Gorinski @xylankant]

Amazing output by my colleagues @fenchri @iiacobacNLP and their intern @chenxi_jw at this year's findings of #EMNLP2022 . Check it out if you're into Code Switching and/or Cross-Lingual Transfer and/or Zero-Shot approaches!
#Huawei #NLProc #naturallanguageprocessing #NLP #AI

[Fenia Christopoulou @fenchri]

Our EntityCS paper w/ @chenxi_jw + @iiacobacNLP, during her internship at Huawei Noah's Ark, is accepted in Findings of #EMNLP2022! :tada:

📄 https://t.co/zPsuz6tAo5

We augment multilingual LMs with entity knowledge by creating an Entity-centric Code-switching corpus! 🧵⬇️

[Philip John Gorinski @xylankant]

Check out this awesome piece of work by my colleagues @fenchri @glampouras_NLP @iiacobacNLP that has made it onto this year's curriculum for #EMNLP2022 !
congrats! #Huawei #NLProc #AI #MachineLearning #Artificial_Intelligence #naturallanguageprocessing #NLP

[Fenia Christopoulou @fenchri]

Happy to share our study on "Training Dynamics for Curriculum Learning" on NLU w/ @glampouras_NLP + @iiacobacNLP to appear at #EMNLP2022!

📄https://t.co/yVXI5YAQgs

Can Training Dynamics as difficulty metrics in CL offer better performance of LMs on mono/cross-lingual NLU?🧵⬇️

[Philip John Gorinski @xylankant]

@theurigeller FINALLY someone's doing what has to be done! This will show 'em who's boss! The crazy boy in his bunker is TREMBLING! Thanks @TheOnion

[Fenia Christopoulou @fenchri]

Presenting our latest work in Huawei Noah’s Ark Lab in collab with Huawei Cloud:
*PanGu-Coder*, a PanGu-α extension, with objectives adapted for text-to-code generation (program synthesis). 317M model variant achieves 17% pass@1 on HumanEval🎉
https://t.co/enlUCm8iTO
👇A thread🧵