bowphs.bsky.social @bowphs.bsky.social · 13/05/2026"How deep is your love? Like the ocean?" Looking out at the Balearic Sea got me thinking about ancient feelings. Just how deep *did* the Romans and Greeks love? Come find out at the first poster session of @LREC2026! 📍 Poster 1196 🕐 11:20 - 13:00 021
Reposted by @bowphs.bsky.socialaicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 14/09/2025The world’s largest NLP conference with almost 2,000 papers presented, ACL 2025 just took place in Vienna! 🎓✨ Here is a quick snapshot of the event via a short interview with one of the authors whose work caught my attention. 🎥 Watch: youtu.be/GBISWggsQOA 032
bowphs.bsky.social @bowphs.bsky.social · 14/09/2025ACL paper: aclanthology.org/2023.acl-lon... Models: github.com/Heidelberg-N... Read more: cl.uni-heidelberg.de/nlpgroup/new... Morphological Analysis Demo: huggingface.co/spaces/bowph... Machine Translation Demo: huggingface.co/spaces/bowphs/ Best Thesis Award: www.gscl.org/en/activitie... 020
bowphs.bsky.social @bowphs.bsky.social · 14/09/2025I am honored to receive the 2025 #GSCL Best Thesis Award at #KONVENS in Hildesheim for my Master’s thesis, which investigates multilinguality and develops language models for Ancient Greek and Latin. Thank you to my mentors and collaborators. I look forward to what comes next. 141
bowphs.bsky.social @bowphs.bsky.social · 27/07/2025Looking at Bruegel's Tower of Babel in Vienna makes you wonder: How can multilingual language models overcome the language barriers? Find out tomorrow! 📍 Level 1 (ironic, right?), Room 1.15-1 🕐 2 PM #ACL2025NLP 030
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025Read the full paper here: arxiv.org/pdf/2506.01629 Reach out if you have any questions or if you are attending ACL and want to say hi. 🙋arxiv.org 020
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025This phenomenon has a visible effect on text generation: In BLOOM-560m, activating 'earthquake' neurons derived from Spanish data at checkpoint 10,000 generates Spanish text. At checkpoint 400,000, the same method yields English text! 120
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025This is not a bug, it's a feature! These layers are repurposing the space to form cross-lingual abstractions. We track this by examining how specific concepts (like "earthquake" or "joy") align across languages. 120
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025We ask a probing classifier: "Given this hidden state from layer l, what is the language of the source text?" The results are striking: earlier checkpoints consistently solve this with high accuracy across layers. Later checkpoints, however, exhibit clear performance drops. 120
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025How and when do multilingual LMs achieve cross-lingual generalization during pre-training? And why do later, supposedly more advanced checkpoints, lose some language identification abilities in the process? Our #ACL2025 paper investigates. 132
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025Read the full paper here: arxiv.org/pdf/2506.01629 Reach out if you have any questions or if you are attending ACL and want to say hi. 🙋arxiv.org 000
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025This phenomenon has a visible effect on text generation: In BLOOM-560m, activating 'earthquake' neurons derived from Spanish data at checkpoint 10,000 generates Spanish text. At checkpoint 400,000, the same method yields English text! 100
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025This is not a bug, it's a feature! These layers are repurposing the space to form cross-lingual abstractions. We track this by examining how specific concepts (like "earthquake" or "joy") align across languages. 100
bowphs.bsky.social @bowphs.bsky.social · 07/06/2025We ask a probing classifier: "Given this hidden state from layer l, what is the language of the source text?" The results are striking: earlier checkpoints consistently solve this with high accuracy across layers. Later checkpoints, however, exhibit clear performance drops. 100
bowphs.bsky.social @bowphs.bsky.social · 06/06/2025Read the full paper here: arxiv.org/pdf/2506.01629 Reach out if you have any questions or if you are attending ACL and want to say hi. 🙋arxiv.org 000
bowphs.bsky.social @bowphs.bsky.social · 06/06/2025This phenomenon has a visible effect on text generation: In BLOOM-560m, activating 'earthquake' neurons derived from Spanish data at checkpoint 10,000 generates Spanish text. At checkpoint 400,000, the same method yields English text! 100
bowphs.bsky.social @bowphs.bsky.social · 06/06/2025This is not a bug, it's a feature! These layers are repurposing the space to form cross-lingual abstractions. We track this by examining how specific concepts (like "earthquake" or "joy") align across languages. 100
bowphs.bsky.social @bowphs.bsky.social · 06/06/2025We ask a probing classifier: "Given this hidden state from layer l, what is the language of the source text?" The results are striking: earlier checkpoints consistently solve this with high accuracy across layers. Later checkpoints, however, exhibit clear performance drops. 100
Reposted by @bowphs.bsky.socialmoritzplenz.bsky.social @moritzplenz.bsky.social · 21/02/2025Debates aren’t always black and white—opposing sides often share common ground. These partial agreements are key for meaningful compromises Presenting “Perspectivized Stance Vectors” (PSVs) — an interpretable method to identify nuanced (dis)agreements 📜 arxiv.org/abs/2502.09644 🧵 More details below 143
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025Read the full paper: aclanthology.org/2025.finding... Work by Creston Brooks, Johannes Haubold, Charlie Cowen-Breen, Jay White, Desmond DeVaul, me, Karthik Narasimhan, and Barbara Graziosiaclanthology.orgAn Annotated Dataset of Errors in Premodern Greek and Baselines for Detecting ThemCreston Brooks, Johannes Haubold, Charlie Cowen-Breen, Jay White, Desmond DeVaul, Frederick Riemenschneider, Karthik R Narasimhan, Barbara Graziosi. Findings of the Association for Computational Lingu... 071
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025Our work brings new computational methods to a field traditionally dominated by manual scholarship, potentially accelerating the discovery of textual errors that have remained hidden for centuries. 130
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025Perhaps most surprising: even powerful models like GPT-4 performed barely above random chance on this specialized task! This highlights the limitations of general-purpose LLMs when dealing with ancient text restoration. 130
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025We tested several error detection methods and found that our discriminator-based approach outperforms all others. Interestingly, scribal errors (the oldest type) are universally more difficult to detect than print or digitization errors across ALL methods. 120
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025Prior work has only evaluated error detection on artificially-generated errors. Our dataset contains REAL errors that naturally accumulated over centuries - the subtle mistakes that survived precisely because they often appear perfectly reasonable. 110
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025Creating this dataset was painstaking! Our domain expert spent over 100 hours reviewing potential errors, categorizing them as scribal errors (from manuscript copying), print errors (from creating editions), or digitization errors (from converting to digital). 120
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025In "An Annotated Dataset of Errors in Premodern Greek and Baselines for Detecting Them," we introduce the first expert-labeled dataset of real errors in ancient texts, enabling proper evaluation of error detection methods on authentic textual problems. 131
bowphs.bsky.social @bowphs.bsky.social · 01/05/2025What did Aristotle actually write? We think we know, but reality is messy. As ancient Greek texts traveled through 2,500 years of history, they were copied and recopied countless times, accumulating subtle errors with each generation. Our new #NAACL2025 paper tackles this fascinating challenge. 1134
