Katie Stasaski

Research Scientist @GoogleDeepMind. Ex-@sfresearch. Ph.D. from UC Berkeley

[Google for Education @GoogleForEdu]

The ultimate #BackToSchool supplies just dropped! 🎒Get ready for smarter learning with new tools in the @GeminiApp! Discover Guided Learning, integrated visuals, and exam prep tools to help students build their knowledge. Learn more: https://t.co/kTEb4gzHb1

[Google @Google]

Here are new @GeminiApp tools to help you learn, understand and study better this school year ✏️

- Guided Learning helps you build a deep understanding of subjects, with step-by-step breakdowns that uncover the “why” and “how”
- Gemini’s responses automatically integrate visuals, like diagrams and YouTube videos, so you can understand more easily
- Flashcards and study guides based on Gemini-generated quizzes or other class materials let you review key concepts

Plus, students can get a Google AI Pro subscription at no cost for 1 year, available in the U.S., Japan, Indonesia, Korea and Brazil.

Learn more about @GeminiApp’s latest updates → https://t.co/IYi4Ns1Nvk

[Salesforce AI Research @SFResearch]

🇨🇦🇨🇦🇨🇦 Welcome to Vancouver! 🇨🇦🇨🇦🇨🇦
13 Paper links below! 👇

The @Salesforce AI Research team brought a baker's dozen AI Research advancements to #NeurIPS2024 this year -- from revolutionizing multimodal agents and time series forecasting to tackling responsible AI evaluation and deployment!

🎯 Attending? Follow us for poster sessions & presentation schedules!
📚 Can't make it? We've curated our complete research collection being showcased this week—bookmark and dive into the work that interests you most!

----

⭐ Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models
📄 https://t.co/jqtKBOT9sy

⭐ INDICT: Code Generation with Internal Dialogues of Critiques for Both Security and Helpfulness
📄 https://t.co/tPIXTwchgr

⭐ MINT-1T: Scaling Open-Source Multimodal Data by 10x: A Multimodal Dataset with One Trillion Tokens
📄 https://t.co/4cYg1TtSkO

⭐ APIGen: An Automated Pipeline for Generating Verifiable and Diversity Function-Calling Datasets
📄 https://t.co/lw1BB8a9A6

⭐ Reverse Transition Kernel: A Flexible Framework to Accelerate Diffusion Inference
📄 https://t.co/MxNWvzFYYj

⭐ Online Iterative Reinforcement Learning from Human Feedback with General Preference Model
📄 https://t.co/3gZC3Ck9BS

⭐ ThinK: Thinner Key Cache by Query-Driven Pruning
📄 https://t.co/4NSZHwh5DZ

⭐Moirai-MoE: Empowering Time Series Foundation Models with Sparse Mixture of Experts
📄 https://t.co/ZTxo6qlySh

⭐ GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluation
📄 https://t.co/WkCuFbQG4h

⭐ UniTST: Effectively Modeling Inter-Series and Intra-Series Dependencies for Multivariate Time Series Forecasting
📄 https://t.co/jSpfevfvZB

⭐ Consent in Crisis: The Rapid Decline of the AI Data Commons
📄 https://t.co/EHlOMDCJzI

⭐ OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
📄 https://t.co/A42mm6sQ7o

⭐ Spider2-V: How Far Are Multimodal Agents From Automating Data Science and Engineering Workflows?
📄 https://t.co/n1k28d5nmS

[Andrew Head @drewmikehead]

Are you a Ph.D. applicant interested in HCI? Consider applying to @PennHCI! This year, my group will have openings that could be great for those interested in HCI/Programming Languages or AI-augmented documents. DMs are welcome!

[Silvio Savarese @silviocinguetta]

How can novel architectures take dev productivity beyond tooling to true human-AI collaboration? CodeGenie's a great start. A new blog outlines the research roadmap for 24/7 pair programmers. #AIResearch #SoftwareEngineeringAI

[Katie Stasaski @katiestasaski]

Excited to share our blog highlighting some of the work I've been involved with at @SFResearch to improve developer productivity.

[Salesforce AI Research @SFResearch]

CodeGenie's impact: 2M+ lines of code accepted, 500K+ chat questions answered. But the real breakthrough? Our vision for a 24/7 AI pair programmer. We’re transforming software engineering, exploring continuous learning and adaptive coding assistance. Check out the technical challenges we're tackling: https://t.co/PrDaLitpgb #FutureOfCoding #Salesforce #AIResearch

[Yingbo @yingbozhou_ai]

Check out our new work that takes advantage of diverse and independent solutions from independent agents

[Salesforce AI Research @SFResearch]

Releasing 🚀 CodeGen2.5 🚀, a small but mighty LLM for code.

- On par with models twice its size
- Trained on 1.5T tokens
- Features fast infill sampling

Blog: https://t.co/8EmiaSrZ2d
Paper: https://t.co/cJlY1trhhJ
Code: https://t.co/FoXGK50TgY
Model: https://t.co/JPUsMDzcmK

[Salesforce AI Research @SFResearch]

🔥Introducing XGen-7B, a new 7B LLM trained on 8K seq. length for 1.5T tokens. Better or comparable results with MPT, Falcon, LLaMA, OpenLLaMA in text & code tasks.
Blog: https://t.co/NFtucUHSMy
Code: https://t.co/d5RjpvCTy4

Training cost ~$150K for 1T token

[Andrew Head @drewmikehead]

My group is admitting Ph.D. students this year! We build tools to help people understand complex things. Recently, we have been really excited about interactive science papers 🧪, math notation ➕, and programs 💻

Reach out with questions, and apply at https://t.co/rRZ5Py0qJH.

[Erik Nijkamp @erik_nijkamp]

🤗@SFResearch at #NeurIPS2022🤗

Meet the team behind CodeGen and see our first steps toward "Conversational Programming".

Join us at the Salesforce booth (#503) for demos and more.

[Uri Alon @urialon1]

There are NLP tasks in which Codex performs better than GPT-3 and T5,
if you convert the NL problem into pseudo-python!

https://t.co/vF5nMcQ8QY
code: https://t.co/PllYtT5MI5
(To appear in #emnlp2022)

Work led by @aman_madaan , with @shuyanzhxyc Yiming Yang @gneubig and me

[Andrew Kuznetsov @andrewkuznet]

Know someone considering a CS/HCI PhD this year?

I keep a list of open research positions to browse!👇
https://t.co/buIsSBA88P

+ if you see a post missing, add it and help others!

Updated daily! Best of luck on your applications, folks!

[Dan Liu @danliu]

Design is hard (II).

A large part of this is choosing colors — *everything* you design needs a color.

I’ve realized that while true mastery takes years, there are simple, scientific rules that can make your colors much nicer.

Here are 10 practical tips for choosing colors.

[Catherine Chen @cathychen23]

If you are a prospective PhD student applying to @Berkeley_EECS, you can submit your application to EAAA for feedback until October 8th 2022 (ten days from today)!

[Ying Xu @YYingXu]

I will be recruiting PhD students for Fall 2023 @UMichEducation. If you are or know of someone interested in research that bridges learning sciences, child development, and human-centered AI, please get in touch with me! Below is a thread about my projects—

[Salesforce AI Research @SFResearch]

(1/3) Meet LAVIS, an open-source DL library for Language-Vision research/applications.

With a unified interface and modular design, LAVIS supports a wide range of tasks, datasets, and state-of-the-art models. Read more > https://t.co/dBL6NGY4Wk

[Katie Stasaski @katiestasaski]

@erik_nijkamp @Replit @amasad Super exciting!

[Erik Nijkamp @erik_nijkamp]

Our model CodeGen is powering @Replit's AI Mode.

Write code with AI assistance.

Impressive work by the folks @Replit @amasad.

https://t.co/b5POgqTOjG
https://t.co/xV0VzFfy3G

[Katie Stasaski @katiestasaski]

@BHedayatnia @Zhou_Yu_AI @tur_gokhan @jaseweston @AlexRudnicky Hi! I am trying to email the organizers group for a question about the workshop but google groups tells me I don't have permission. Is there another email address I should use? Thank you!