Auton Lab, Carnegie Mellon retweeted
Huge congratulations to my amazing collaborators Yifu Cai, @Arjun_7m, and Dr. Artur Dubrawski at @AutonLab @CMU_Robotics, @SCSatCMU
Auton Lab, Carnegie Mellon retweeted
Frontier models like Gemini excel in areas like code and image understanding. But how well do they handle time series reasoning?
(Re-)Introducing TimeSeriesExam, which just won the NeurIPS'24 TSALM Best Paper Honorable Mention🏆
The Gemini 2.0 era is here. And we’re excited for you to start building with it.
A quick rewind of what we just released ⏪
Gemini 2.0 Flash ⚡ comes with low latency and better performance.
🔵 You can now access an experimental version in @GeminiApp on the web, while Gemini Advanced users can try Deep Research, a new AI research assistant.
🔵 Developers can begin building through the Gemini API in @Google AI Studio and Vertex AI
2.0 is also enabling new research prototypes of AI agents, including:
🔵 Project Astra, which explores future capabilities of a universal AI assistant
🔵 Project Mariner, which shows what’s possible for human-agent interaction, starting with your browser
🔵 Jules, an experimental AI-powered coding agent
Finally, we’re exploring how 2.0 can be used in agents across domains — from navigating the virtual world of video games to applying its spatial reasoning capabilities to robotics. 🤖
goo.gle/gemini-2
Auton Lab, Carnegie Mellon retweeted
Excited to share TimeSeriesExam for systematic evaluation of time series reasoning capabilities of LLMs. Think your LLM can reason on time series concepts? Take it for a spin on the TimeSeriesExam! Now publicly available on HuggingFace :)
Auton Lab, Carnegie Mellon retweeted
Heartiest congratulations to our amazing team!
I am applying for PhD positions for Fall 2025. If you see a fit with your lab or interests, or just want to brainstorm on ideas, please DM :)
Huge congratulations to my amazing collaborators Yifu Cai, @Arjun_7m, and Dr. Artur Dubrawski at @AutonLab @CMU_Robotics, @SCSatCMU
Auton Lab, Carnegie Mellon retweeted
Time Series & Language: A Powerful Pair 📈📉
Loved seeing some recent work time series & text work! It reminded me of our very own work JoLT, one of the first time series and text models !
Want to learn more? Check out JoLT, and our latest paper, on the TimeSeriesExam!
ALT An overview of JoLT. Given a time-series and an optional textual prompt as input, JoLT produces text as output. We pre-train a Transformer using the masked time-series reconstruction objective to use as an encoder, and the pre-trained language model as the decoder. The Q-Former is trained to align time-series and text representations. Learnable query tokens are used to extract time-series features conditioned on textual prompts.
ALT Each template in the TimeSeriesExam evaluates a specific category, and includes a question, list of options, example question and answer pair for in-context learning, and optionally a hint and descriptions of complicated technical terms. Here, GPT-4o showcases its ability to transfer visual understanding and time series concepts into effective reasoning.
Auton Lab, Carnegie Mellon retweeted
#ICML2024 #ICML
Our new work provides a highly effective and scalable method for enriching deep neural networks to decide whether it has sufficient evidence for reaching a reliable prediction.
Paper 📒 : huggingface.co/papers/2406.1…
Code 💻: github.com/ifiaposto/Distanc…
Poster 📸: icml.cc/virtual/2024/poster/…
Details Below 🧵
Auton Lab, Carnegie Mellon retweeted
Thrilled to announce @SkildAI! Over the past year, @gupta_abhinav_ and I have been working with our top-tier team to build an AI foundation model grounded in the physical world. Today, we’re taking Skild AI out of stealth with $300M in Series A funding:
forbes.com/sites/rashishriva…
Auton Lab, Carnegie Mellon retweeted
We’re striving to build the most scalable, robust, and generalizable robotics foundation model. Thank you to our team for all their hard work so far & our investors for believing in our mission. Follow along with us @SkildAI for the latest. Blog: skild.ai/blogs/announcing-ou…
Auton Lab, Carnegie Mellon retweeted
ICML’24:
“A Rate-Distortion View of Uncertainty Quantification”:
arxiv.org/abs/2406.10775
Distance Aware Bottleneck is a method for enriching deep neural networks to decide whether it has sufficient evidence for reaching a reliable prediction.
@ifaposto @ben_eysenbach Dubrawski
Auton Lab, Carnegie Mellon retweeted
The Auton Lab has proposed MOMENT: one of the first families of open-source foundation models for time series analysis!
The models serve as a building block for diverse tasks such as forecasting, classification, & anomaly detection.
Learn more: moment-timeseries-foundation…
Auton Lab, Carnegie Mellon retweeted
Exciting news! Just as foundation models revolutionized text and image modeling, it's now time for Time Series Data! Introducing MOMENT, a family of Open Time Series Foundation Models, freshly accepted at @icmlconf 2024! 🎉
Auton Lab, Carnegie Mellon retweeted
Excited to present the first day of the @RealAAAI 2024 Spring Symposium on Clinical Foundation Models! 🎊 Today's lineup includes keynotes, a panel discussion, demos, contributed talks, and a poster session!
@clinicalfms
Auton Lab, Carnegie Mellon retweeted
BOOM 💥 The competition in LLM time series space is heating up.
Lag-Llama was making rounds yesterday and some people claimed it was “the first” open source LLM. But it turns out on 6 February before Lag Llama Carnegie Mellon released their own open source LLM “Moment”. 🔥🔥🔥🔥🔥
So MOMENT, not Lag Llama is THE FIRST open source time series LLM. 🚀🚀🚀🚀🚀
“MOMENT: A FAMILY OF OPEN TIME-SERIES FOUNDATION MODELS”
#llm
#timeseries
Auton Lab, Carnegie Mellon retweeted
🤖 We have decided to extend the deadline for our @RealAAAI Spring Symposium on Clinical Foundation Models @clinicalfms to Jan 26, 2024 🥳! CfP, timelines, attendance info available at clinicalfoundationmodels.git… #ML4H #MLHC
🎉Exciting News! Join us for the AAAI-24 Spring Symposium on Clinical Foundation Models! 🚀Explore captivating talks, engaging panels, & cutting-edge demos. All details (CfP, timelines, ...) available at 👉clinicalfoundationmodels.git…. Hurry, papers due by Jan 5, 2024 @clinicalfms
Auton Lab, Carnegie Mellon retweeted
🎉Exciting News! Join us for the AAAI-24 Spring Symposium on Clinical Foundation Models! 🚀Explore captivating talks, engaging panels, & cutting-edge demos. All details (CfP, timelines, ...) available at 👉clinicalfoundationmodels.git…. Hurry, papers due by Jan 5, 2024 @clinicalfms
Auton Lab, Carnegie Mellon retweeted
Tired of reading about Q*?
Want to instead learn about improving the quality of labels in your data?
Then check out our recent NeurIPS D&B paper on benchmarking label error detection tools with AQuA: arxiv.org/abs/2306.09467.
3 reasons why you should check out AQuA: (1/5)
ALT Labeling errors in widely used benchmarks: CIFAR-10, Clothing-100k, MIT-BIH and TweetEval Hate Speech datasets.
Auton Lab, Carnegie Mellon retweeted
"Near-optimal Policy Identification in Active Reinforcement Learning" (top 5% paper)
Xiang and Viraj are presenting this work at @iclr_conf on Monday (Poster session 2 MH1-2-3-4 #101)
Presentation: Oral 2, track 4: Reinforcement Learning
🔖 Bookmark here: iclr.cc/virtual/2023/poster/…
Auton Lab, Carnegie Mellon retweeted
We lost a colleague, dedicated member of the RI Community and a friend. Thank you, Alan. ri.cmu.edu/obituary-alan-gui…
Auton Lab, Carnegie Mellon retweeted
1/ Machine Learning is extensively applied to "phenotype" clinical data, BUT these models often seek answers to the WRONG CLINICAL question.
For a Phenotype to be ACTIONABLE, it should discover patients that benefit (harmed) from a therapy!
#ACC23
autonlab.org/auton-survival/…
Auton Lab, Carnegie Mellon retweeted
1/ Most time-series anomaly detection models don't need labels for training. So why should we need labels to select good models?
In our accepted #ICLR paper (spotlight), we show how we can select good anomaly detection models given unlabeled data and a set of candidate models.
ALT The Model Selection Workflow. We identify three classes of surrogate metrics of model quality, and propose a novel robust rank aggregation framework to combine multiple rankings from metrics.