Lütfi Kerem Şenel

NLP Researcher / PhD Candidate at @CisLmu / @LMU_Muenchen supervised by @HinrichSchuetze

[Ibna Mattuta @swahiligal]

What makes the earthquakes in Turkey one of the worst disaster of the century?

A basic explanation for those who still don't fully understand

[Lütfi Kerem Şenel @lksenel]

Performance on CoDA21 improves linearly with model size. However, even the largest 170B parameter model performs still poorly (<50% acc.), while expert humans can solve the task perfectly. Solving CoDA21 takes time and is not really well suited for crowd-sourced evaluation.

[Lütfi Kerem Şenel @lksenel]

The paper for BIG-bench🪑is finally live at https://t.co/ShujvO3yzK. It contains 204 diverse and challenging language tasks, including our context definition alignment task, https://t.co/r52ZRcwOSg.

A nice summary thread for BIG-bench by @jaschasd:
https://t.co/iQgiyI0jMP

[Jascha Sohl-Dickstein @jaschasd]

After 2 years of work by 442 contributors across 132 institutions, I am thrilled to announce that the https://t.co/wezEGzDEHt paper is now live: https://t.co/4Yg36EB9Ru. BIG-bench consists of 204 diverse tasks to measure and extrapolate the capabilities of large language models.

[Jascha Sohl-Dickstein @jaschasd]

After 2 years of work by 442 contributors across 132 institutions, I am thrilled to announce that the https://t.co/wezEGzDEHt paper is now live: https://t.co/4Yg36EB9Ru. BIG-bench consists of 204 diverse tasks to measure and extrapolate the capabilities of large language models.

[Silvia Severini @silviaseverini3]

Excited to share that our paper “Towards a Broad Coverage Named Entity Resource: A Data-Efficient Approach for Many Diverse Languages" has been accepted at #LREC2022! 🎉
Joint work with @imani_ayyoob, @PDufter, and @HinrichSchuetze. [1/3]

Preprint: https://t.co/sX9GR3G0eU

[Antonis Maronikolakis @antmarakis]

Our work (w/ @HinrichSchuetze and @4digitaldignity) collecting #HateSpeech data in collaboration with affected communities around the globe has been accepted to #ACL2022 Findings. [1/3]

⚙️ https://t.co/nFCzUv5P9c
📝 https://t.co/O12ddiouj7

[Aykut Koc Lab - Bilkent @KocLab_Bilkent]

1- In collaboration with Schütze Lab (LMU), Icon Lab (Bilkent) and ASELSAN Research Center, KocLab has published "Learning Interpretable Word Embeddings via Bidirectional Alignment of Dimensions with Semantic Concepts" in Information Processing and Management journal. The paper-

[Lütfi Kerem Şenel @lksenel]

7/ CoDA21 is publicly available at https://t.co/h9vIbG7ObO
Do you have a cool model with strong NLU? Go ahead and test it on CoDA21 to see if it can compete with humans.

Looking forward to meeting you at #ACL2022, hopefully in person.

[Lütfi Kerem Şenel @lksenel]

I am thrilled to share that our paper "CoDA21: Evaluating Language Understanding Capabilities of NLP Models With Context-Definition Alignment
" got accepted to #ACL2022. Joint work with @timo_schick and Hinrich Schütze.

https://t.co/v1enLGPk4e

Below is a summary thread👇

[Lütfi Kerem Şenel @lksenel]

6/ Performing well on CoDA21
requires detailed understanding of contexts, performing complex inference, and having world
knowledge, which are crucial skills for NLP. Our results indicate a lack of these skills in the current state of the art in NLP.

[Masoud Jalili @_masoudjalili]

Check our new paper at #EMNLP2021 "Graph Algorithms for Multiparallel Word Alignment". We get better results compared to bilingual aligners even when we use noisy translations.
Joint work: @imani_ayyoob
@lksenel @PDufter @HinrichSchuetze
PDF: https://t.co/rCoOvgV8dj
#NLProc

[Timo Schick @timo_schick]

🎉 Happy to share that "It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners" (w/@HinrichSchuetze) has been accepted to #NAACL2021 #NLProc 🥳

📷 Camera-ready: https://t.co/73RPkN2FuD
🖥️ Code: https://t.co/YCrBnggof7
📝 Blog: https://t.co/YdCgsopk7w

[Alexandra Chronopoulou @alexandraxron]

Happy to share that our paper "Improving the Lexical Ability of Pretrained Language Models for Unsupervised Neural Machine Translation" got into #NAACL2021!🎉Joint work w/ @StDario1 and Alex Fraser.

Here is a summary thread👇

[Timo Schick @timo_schick]

🎉New paper🎉 In "Self-Diagnosis and Self-Debiasing", we investigate whether pretrained LMs can use their internal knowledge to discard undesired behaviors and reduce biases in their own outputs (w/@4digitaldignity + @HinrichSchuetze) #NLProc

Link: https://t.co/PQ04Ly3eFd [1/3]

[Lütfi Kerem Şenel @lksenel]

I am happy to share that our work on evaluating word understanding of pretrained language models using dictionary definitions was accepted at EACL 2021.
Joint work with @HinrichSchuetze

Preprint: https://t.co/VMSvWA61wg