I am happy to share that I defended my PhD thesis this week. I want to thank all the people I collaborated with and met at conferences and workshops over the years! Especially a big thank you to my PhD supervisor, Matthias Hagen, and to Laura Dietz and Udo Kruschwitz for reviewing my thesis.
Preprint: arxiv.org/abs/2503.09701
#NLP #ActiveLearning #LLMs
Active learning is practical, so much of it isn't documented. We surveyed the NLP community to collect insights on active learning.
Some key takeaways:
- Data annotation is still a bottleneck.
- LLMs complement rather than replace careful annotation.
- Longstanding challenges limit wider adoption.
🎉 Somewhat late, but excited to share: our paper “Reassessing Active Learning Adoption in Contemporary NLP: A Community Survey” has been accepted to #EACL2026!
We asked: What are data annotation needs in the era of LLMs? How is active learning actually used? How does it compare to other methods?
Wow.
Twitch says they won't allow "hatred, prejudice, or intolerance" on their platform.
But their automated moderation tool (AutoMod) really sucks
It misses ≈94% of hateful messages (unless there are slurs)
+ it blocks ≈90% of benign messages that happen to use sensitive words in positive ways.
⚖️ Measuring Scalar Constructs in Social Science with LLMs
with rising (and established) stars in Computational Social Science
@haukelicht.bsky.social
@rupak-s.bsky.social
@patrickwu.bsky.social
@pranavgoel.bsky.social
@elliottash.bsky.social
@alexanderhoyle.bsky.social
arxiv.org/abs/2509.03116
We just released "German Commons", the largest openly-licensed German text dataset for LLM training: 154B tokens with clear usage rights for research and commercial use.
huggingface.co/datasets/coral-nlp/german-commons
Thank you for the clarification, however, KL3M was released in February 2024. That being said, I appreciate your contribution, this is solid work.
There is really no need to exaggerate your claim, this undermines the good work you have done.
Still, congrats on the publication!
Are you sure that you are the first? What about the KL3M models?
arxiv.org/pdf/2504.07854
🏆 Thrilled to share that our HateDay paper has received an Outstanding Paper Award at #ACL2025
Big thanks to my wonderful co-authors: @deeliu97.bsky.social, Niyati, @computermacgyver.bsky.social, Sam, Victor, and @paul-rottger.bsky.social!
Thread 👇and data avail at huggingface.co/datasets/man...
Congratulations, Manuel!
Honored to win the ICTIR Best Paper Honorable Mention Award for "Axioms for Retrieval-Augmented Generation"!
Our new axioms are integrated with ir_axioms: github.com/webis-de/ir_...
Nice to see axiomatic IR gaining momentum.
Happy to share that our paper "The Viability of Crowdsourcing for RAG Evaluation" received the Best Paper Honourable Mention at #SIGIR2025! Very grateful to the community for recognizing our work on improving RAG evaluation.
📄 webis.de/publications...
Dory from finding nemo with the quote: "I remember it like it was yesterday. Of course, I dont remember yesterday."
Do not forget to participate in the #TREC2025 Tip-of-the-Tongue (ToT) Track :)
The corpus and baselines (with run files) are now available and easily accessible via the ir_datasets API and the HuggingFace Datasets API.
More details are available at: trec-tot.github.io/guidelines
Oh no, what happened to Argilla? @hf.co Could you explain what's going on? It has barely been a year since you bought it.
#nlproc #nlp #ml
@ai2.bsky.social Any plans for plagiarism detection in semantic scholar? This would be incredibly useful, especially with the growing influx of (semi-)automatically generated papers.
Big fan of @ai2.bsky.social's semantic scholar feeds. Usually great for paper recommendations. Yesterday it recommended... a paper that blatantly plagiarized from a former student's thesis that I co-supervised. So, I guess the algorithm really knows my interests 😅.
Our recent paper on the impact of register (genre) on LLM performance. Key points: news do poor in evaluation, while opinionated texts are among the best. We hope this work can be used to understand the impact of register on LLMs and improve training data mixes! arxiv.org/abs/2504.01542
Plot shows the relationship between compute used to predict a ranking of datasets and how accurately that ranking reflects performance at the target (1B) scale of models pretrained from scratch on those datasets.
Ever wonder how LLM developers choose their pretraining data? It’s not guesswork— all AI labs create small-scale models as experiments, but the models and their data are rarely shared.
DataDecide opens up the process: 1,050 models, 30k checkpoints, 25 datasets & 10 benchmarks 🧵
A bit of a mess around the conflict of COLM with the ARR (and to lesser degree ICML) reviews release. We feel this is creating a lot of pressure and uncertainty. So, we are pushing our deadlines:
Abstracts due March 22 AoE (+48hr)
Full papers due March 28 AoE (+24hr)
Plz RT 🙏
Can a Large Language Model (LLM) with zero Pokémon-specific training achieve expert-level performance in competitive Pokémon battles?
Introducing PokéChamp, our minimax LLM agent that reaches top 30%-10% human-level Elo on Pokémon Showdown!
New paper on arXiv and code on github!
(1/8) Excited to share some new work: TESS 2!
TESS 2 is an instruction-tuned diffusion LM that can perform close to AR counterparts for general QA tasks, trained by adapting from an existing pretrained AR model.
📜 Paper: arxiv.org/abs/2502.13917
🤖 Demo: huggingface.co/spaces/hamis...
More below ⬇️
After 6+ months in the making and over a year of GPU compute, we're excited to release the "Ultra-Scale Playbook": hf.co/spaces/nanot...
A book to learn all about 5D parallelism, ZeRO, CUDA kernels, how/why overlap compute & coms with theory, motivation, interactive plots and 4000+ experiments!
More than 8500 submissions to ACL 2025 (ARR February 2025 cycle)! That is an increase of 3000 submissions compared to ACL 2024. It will be a fun reviewing period. 😅💯
@aclmeeting.bsky.social #ACL2025 #ACL2025nlp #NLP
Fixed: We need your support *for a* web survey.
Sorry, it seems bluesky has no edit feature yet.
I have the feeling, I did not reach the NLP crowd on bluesky yet. Where are the large groups here? Who do I have to ping❓
Please consider participating or sharing our survey! (If you have any experience with supervised learning in natural language processing, you are eligible to participate in our survey.)
The survey has a partial focus on, but not is limited to, active learning. See the original post for details.
➡️ Extended Deadline: January 26th, 2025.
🔥 𝐅𝐢𝐧𝐚𝐥 𝐂𝐚𝐥𝐥 𝐚𝐧𝐝 𝐃𝐞𝐚𝐝𝐥𝐢𝐧𝐞 𝐄𝐱𝐭𝐞𝐧𝐬𝐢𝐨𝐧: Survey on Data Annotation and Active Learning
We need your support in web survey in which we investigate how recent advancements in NLP, particularly LLMs, have influenced the need for labeled data in supervised machine learning.
#NLP #NLProc #ML #AI
Hallo and happy New Year #NLProc :) Julia Romberg, a postdoc in my group in Cologne, together with other collaborators, is conducting a survey on the use of Active Learning in NLP. Find the link in the thread below!
