Exciting news: the #PyConUS 2027 website is live, and our Call for Proposals is officially open!! 🎉
We invite you to share your work with the Python community in Long Beach, CA 🐍
🗓️ Submit your proposal by December 19, 2026.
All the details: pycon.blogspot.com/2026/11/p…
Liling Tan retweeted
As we close out 2026, the PSF is running our annual fundraiser with an important message: Python is for everyone 💛🐍🌏💙
Your support helps keep #Python open & thriving for everyone, including our 20 incredible Fiscal Sponsorees!
Donate today 👉 linktr.ee/thepsf
#PythonForEveryone
Liling Tan retweeted
If you are preparing for interview with post training team, going through Nathan’s book is a good use of your time rlhfbook.com
I regularly hear from people that my post-training book/writing has helped them get a job at labs by focusing on material that develops intuitions for how training methods work. It's very validating, because this type of "idea/math heavy" writing doesn't really sell many books. It's rewarding the right people who put the effort into learning. I'm happy I made the time for it.
Liling Tan retweeted
Know a pro AI & Security researcher who loves security and agent/AI software? Our NVIDIA team is growing and we have exciting roles open for scientists who want to work hands-on in charting the course to the future.
NVIDIA builds tools that help the world secure AI and use AI for security - join us if you'd like to be part of that story.
Positions are based in US, including NVIDIAs incredible Santa Clara headquarters, or elsewhere. Looking for researchers with a track record in both AI and in Security. See link in comments to apply today.
Liling Tan retweeted
Mamba-2 doesn’t need most of its parts, three components carry nearly all of its accuracy.
Mamba-2S (“2Mamba2Furious”, Mongaras & Larson) keeps only a softplus decay mask, a width-2 conv, and dt discretization, dropping the Z gate, D residual and SiLU.
arxiv.org/abs/2602.17363
Liling Tan retweeted
We worked on a llama.cpp impl for DS4.1-Flash + made Unsloth dynamic GGUFs for it!
The engram / PLE can be swapped out from MXFP8 to Q6_K to Q5_K and IQ4_NL precision - we found Q6_K approx = MXFP8 and IQ4_NL loses a tiny bit of accuracy.
The engram can stream fully from disk!
DeepSeek V4.1-Flash-1017 can now run locally! 🐳
We made 1-bit retain ~88.2% accuracy while shrinking size from 508GB to 250GB (−52%).
Run on a 128GB+ RAM device.
V4.1 Flash 1017 is the strongest open model to date
Guide: unsloth.ai/docs/models/deeps…
GGUF: huggingface.co/unsloth/DeepS…
Liling Tan retweeted
If you're building a coding agent harness, give your agent this paper and let it figure out how to improve itself.
arxiv.org/pdf/2609.20804
Liling Tan retweeted
You can now train your own Decision model with our free notebook! 💡
Qwen3.5-4B will generate decisions instead of text on just 8GB VRAM locally.
Learn to data prep (state, questions, gold answers), train, serve.
Notebook: colab.research.google.com/gi…
Guide: unsloth.ai/docs/basics/train…
You can now train your own Decision model like Jev locally!
We increased Qwen3.5 0.8B’s aggregate accuracy from 20.7% to 74.3% across 3 decision benchmarks - on just 4GB VRAM.
Turn any LLM like Qwen3.8, Gemma 4 into decision models with our open-source Unsloth repo.
We fine-tuned with a Clef head using Unsloth and LoRA (r=64) for one epoch, increasing downstream accuracy from 30–37% to 78%.
GitHub: github.com/unslothai/unsloth
Guide and Notebooks: unsloth.ai/docs/basics/train…
Liling Tan retweeted
We made it possible to train your own Decision model locally on just 3GB VRAM!
We converted Qwen, Gemma, Llama all into decision models by fine-tuning using a Clef head, boosting accuracy from 30 to up to 78%.
You can try it yourself via Unsloth Desktop!
You can now train your own Decision model like Jev locally!
We increased Qwen3.5 0.8B’s aggregate accuracy from 20.7% to 74.3% across 3 decision benchmarks - on just 4GB VRAM.
Turn any LLM like Qwen3.8, Gemma 4 into decision models with our open-source Unsloth repo.
We fine-tuned with a Clef head using Unsloth and LoRA (r=64) for one epoch, increasing downstream accuracy from 30–37% to 78%.
GitHub: github.com/unslothai/unsloth
Guide and Notebooks: unsloth.ai/docs/basics/train…
Liling Tan retweeted
Quality filters that work in English degrade measurably in every other language you cover. Tested across 35 languages.
The vendor question is not how many languages they cover. It is how quality is measured per language pair.
pangeanic.com/data-for-ai/ev…
Liling Tan retweeted
Unsloth Desktop can serve local Jev type decision models!
We made a real time packing demo powered by local Laya through Unsloth’s Decision API - suitcase items update as you type and change your travel plans!
More optims coming soon to make it even faster for local hardware!
You can now run Laya Decision models locally on just 4GB RAM! 🔥
Works on CPU, Mac, Windows, Linux and GPU setups.
Serve Laya through a Jev-compatible API via Unsloth Desktop.
GitHub: github.com/unslothai/unsloth
Guide: unsloth.ai/docs/models/decis…
Liling Tan retweeted
PokéAgent Challenge was accepted to NeurIPS 2026!
This has been a multi-year effort toward standardized Pokémon benchmarks for agents, starting with PokéChamp, our ICML 2025 Spotlight, then the NeurIPS 2025 competition, and now a much broader eval + dataset effort.
Huge thanks to everyone who contributed to the retrospective :)
Liling Tan retweeted
📣 Second Call for Papers: ComputEL-10
We invite submissions for ComputEL-10, taking place March 2–3, 2027, in Honolulu, Hawaiʻi, USA (hybrid).
📅 Submission deadline: October 2, 2026
🔗 Submit paper: softconf.com/p/ComputEL2027
For more info: computel-workshop.org/comput…
Liling Tan retweeted
Looking for people who are qualified and willing to #review submissions about:
1. Modelling Dialect Variation in Low-Resource Machine Translation
2. Image Captioning for a Low-Resource Language (Encoder–Decoder Architecture)
#ACM #TALLIP #NLP #linguistics #LowResource #MT
We’re teaching a new course at Penn: CIS 6280 · World Models 🌍
cis.upenn.edu/~cis6280/
Every day, LLMs amaze me with one more thing they can solve. But I’ve always felt there’s more to learn than what humans have written down—from observing the world, interacting with it, and experiencing what happens next. To me, that’s what world models are about—and why I see them as one of AI’s next big bets.
However, I struggled to find a systematic course on them—so we are building one!
Topics will include: representation learning, generative models, simulation, model-based RL, video and 3D generation, world models for robotics, reasoning, and code-based world models.
Hands-on work will include:
- Building an environment
- Training a world model
- Learning a policy
and a final research project with leaderboard.
We’re already six lectures in. Slides, demos, and readings are public, and we’ll keep adding materials throughout the fall.
Big thanks to our TA team— @TongMutianTMT @hagsaeng_bag @EnxinSong @KeelyAi04 —for helping bring this course to life!
Liling Tan retweeted
The most logical path to safe AI is through open and transparent AIs.
Liling Tan retweeted
Come work with us to improve multilinguality for AFMs!
jobs.apple.com/en-us/details…
Liling Tan retweeted
Checkout the latest XGBoost update "Introducing the XGBoost Vector-Leaf Model" from jiamin yuan and Rory Mitchell
xgboost.ai/2026/08/25/introd…
































