Daniel Varab retweeted
Did not include it in the SYNTH paper but we did some jspace analysis of Baguettotron and the commit zone topology is radically different from Gemma-270m (+very high level of connection between features). Looks to be more an effect of synth pretraining than deep arch.
We're hiring Rust developers!
jobs.ashbyhq.com/dottxt
[LOOKING FOR EMERGENCY REVIEWERS]
This ARR cycle has been brutal on the reviewing system. I'm still short reviewers for 4 papers in the are of Language Modeling. Drop me a message if you're an NLP researcher who could take even one emergency review - I'd be incredibly grateful🙏
Daniel Varab retweeted
I am hiring: candidate.hr-manager.net/App… So if you are interested in training/evaluating LMs and language/linguistics, consider applying!
Daniel Varab retweeted
Reminder: EACL 2026 submissions (via ACL Rolling Review) are due today, October 6, AoE!
Do not forget to leave enough time for filling in the responsible NLP checklist! :)
Submission link: openreview.net/group?id=aclw…
Daniel Varab retweeted
The CIS institute at LMU Munich has an opening for a tenure-track Akademische/r Rat/Rätin (roughly equivalent to lecturer or junior professor) in NLP / Computational Linguistics.
Details are available here: cis.lmu.de/web/arposition202…
Daniel Varab retweeted
🕊️ Lifetime Achievement Award at #ACL2025NLP
A standing ovation for Prof. Kathy McKeown, recipient of the ACL 2025 Lifetime Achievement Award! 🌟
Daniel Varab retweeted
Workaccount2 on Hacker News just coined the term "context rot" to describe the thing where the quality of an LLM conversation drops as the context fills up with accumulated distractions and dead ends news.ycombinator.com/item?id…

ALT Comment by Workaccount2, 9 hours ago: They poison their own context. Maybe you can call it context rot, where as context grows and especially if it grows with lots of distractions and dead ends, the output quality falls off rapidly. Even with good context the rot will start to become apparent around 100k tokens (with Gemini 2.5). They really need to figure out a way to delete or "forget" prior context, so the user or even the model can go back and prune poisonous tokens. Right now I work around it by regularly making summaries of instances, and then spinning up a new instance with fresh context and feed in the summary of the previous instance.
Daniel Varab retweeted
📢 #ACL2025NLP Dear community! The reviewing deadline for ACL 2025 is approaching -- 23 March! We received over 8000 submissions, the highest number in ACL history. It is particularly important that you submit your reviews on time. A big thank you in advance for all your efforts!
Daniel Varab retweeted
Today, we release QwQ-32B, our new reasoning model with only 32 billion parameters that rivals cutting-edge reasoning model, e.g., DeepSeek-R1.
Blog: qwenlm.github.io/blog/qwq-32…
HF: huggingface.co/Qwen/QwQ-32B
ModelScope: modelscope.cn/models/Qwen/Qw…
Demo: huggingface.co/spaces/Qwen/Q…
Qwen Chat: chat.qwen.ai
This time, we investigate recipes for scaling RL and have achieved some impressive results based on our Qwen2.5-32B. We find that RL training con continuously improve the performance especially in math and coding, and we observe that the continous scaling of RL can help a medium-size model achieve competitieve performance against gigantic MoE model. Feel free to chat with our new models and provide us feedback!
Daniel Varab retweeted
Good to see @EU_Commission promoting OS LLMs in Europe. However (1) "OpenEuroLLM" is appropriating a name (#EuroLLM) which already exists, (2) it is certainly *not* the "first family of open-source LLMs covering all EU languages" 🧵
Daniel Varab retweeted
Excited to share we are hiring for a new type of scholar role.
Algorithm interface scholars will re-imagine how humans interact with algorithms.
These scholars will lead the frontier in UI-algorithm co-design. 🔥
Join us at @CohereForAI @cohere ✨
jobs.ashbyhq.com/cohere/dc7d…
Daniel Varab retweeted
It's here, and it's MIT! 👀huggingface.co/microsoft/phi…
Daniel Varab retweeted
📣 I am looking for a new Ph.D. student! 📣
The position is part of a @DASHHPhD project on (multimodal) #LLMs for laser operations with me (@DS_research_HH @unihh) and @HTuennermann from @desy. Deadline: October 20th! Feel free to reach out! 🚀🛸💥
dashh.org/application/phd_to…
Daniel Varab retweeted
Trying to get faster finetuning? How about 𝟏𝟎𝐱 𝐟𝐚𝐬𝐭𝐞𝐫? Our new method, Fast Forward, will help you trim an order of magnitude off your already-efficient LoRA or DoRA finetuning by using … line search, the most basic optimization algorithm.
Daniel Varab retweeted
All @COLM_conf accepted papers are now online, including reviews, decision text, and discussions:
colmweb.org/AcceptedPapers.h…
We still have a limited number of tickets left:
whova.com/portal/registratio…
Daniel Varab retweeted
Keeping Indigenous languages strong in northern Australia: a PhD opportunity at Charles Darwin University, under First Nations leadership. We have a $40k/year domestic scholarship. Please send expressions of interest to [email protected].
Daniel Varab retweeted
Still one week to go! 🚀 Please share this with anyone who may be interested in a 3-year postdoc with an amazing NLP team!
Exciting job opportunity! 🧑🔬 We at the School of Informatics, University of Edinburgh (@InfAtEd, @EdinburghNLP) are looking for a Research Associate (~postdoc) in NLP and ML for a 3-year project fully funded by @Huawei via the Huawei-Edinburgh Joint Lab initiative!
The project focuses on improving the robustness and trustworthiness of LLMs when answering complex questions and solving complex tasks 🚀
📅 Apply by 15th July 2024
🔗 Link to apply: elxw.fa.em3.oraclecloud.com/…
#NLP #MachineLearning #ResearchJobs #AcademicTwitter


























