Cathy Wu
@cathywu.bsky.social
AI & Transportation | MIT Associate Professor
Interests: AI for good, sociotechnical systems, machine learning, optimization, reinforcement learning, public policy, gov tech, open science.
Science is messy and beautiful.
http://www.wucathy.com
Interests: AI for good, sociotechnical systems, machine learning, optimization, reinforcement learning, public policy, gov tech, open science.
Science is messy and beautiful.
http://www.wucathy.com
Pinned
Cathy Wu
@cathywu.bsky.social
· Aug 11
This project was 4 years in the making and it's finally out!
We found that controlling vehicle speeds to mitigate traffic across a city can cut carbon emissions between 11 and 22 percent. To do this, we used deep reinforcement learning to optimize one million eco-driving scenarios. 🚗🤖🧠
We found that controlling vehicle speeds to mitigate traffic across a city can cut carbon emissions between 11 and 22 percent. To do this, we used deep reinforcement learning to optimize one million eco-driving scenarios. 🚗🤖🧠
Reposted by Cathy Wu
I'm gearing up for next week's Hype Studies conference in Barcelona, so I'm thinking about this stuff a lot. @davekarpf.bsky.social makes some great points here, but there's a 'Yes and...'
Tesla's Roadmap to Nowhere
I'll give him this: Elon Musk has a solid grasp of futurity-as-PR.
davekarpf.substack.com
September 5, 2025 at 12:15 PM
I'm gearing up for next week's Hype Studies conference in Barcelona, so I'm thinking about this stuff a lot. @davekarpf.bsky.social makes some great points here, but there's a 'Yes and...'
We should be asking ourselves more what research should be done vs what research can be done
"Can Large Language Models" returns 8k+ hits on Google Scholar.
"Should Large Language Models" returns 73 hits.
Before asking can...?, ask should...? and you'll save yourself a year's worth of research in some cases.
"Should Large Language Models" returns 73 hits.
Before asking can...?, ask should...? and you'll save yourself a year's worth of research in some cases.
August 17, 2025 at 11:00 PM
We should be asking ourselves more what research should be done vs what research can be done
🔥 New workshop at @neuripsconf.bsky.social!
DiffCoALG bridges the gap between classic algorithms & differentiable learning.
Think: LLM reasoning, routing, SAT, MIP — neurally optimized.
📌 Submit by Aug 22! 🤖🧠
🔗 sites.google.com/view/diffcoa...
#NeurIPS2025
DiffCoALG bridges the gap between classic algorithms & differentiable learning.
Think: LLM reasoning, routing, SAT, MIP — neurally optimized.
📌 Submit by Aug 22! 🤖🧠
🔗 sites.google.com/view/diffcoa...
#NeurIPS2025
DiffCoALG@NeurIPS 2025
About this Workshop
Combinatorial algorithms are fundamental across a wide range of domains, owing to their ability to model optimization and decision-making tasks under complex constraints. These alg...
sites.google.com
August 12, 2025 at 2:16 PM
🔥 New workshop at @neuripsconf.bsky.social!
DiffCoALG bridges the gap between classic algorithms & differentiable learning.
Think: LLM reasoning, routing, SAT, MIP — neurally optimized.
📌 Submit by Aug 22! 🤖🧠
🔗 sites.google.com/view/diffcoa...
#NeurIPS2025
DiffCoALG bridges the gap between classic algorithms & differentiable learning.
Think: LLM reasoning, routing, SAT, MIP — neurally optimized.
📌 Submit by Aug 22! 🤖🧠
🔗 sites.google.com/view/diffcoa...
#NeurIPS2025
This project was 4 years in the making and it's finally out!
We found that controlling vehicle speeds to mitigate traffic across a city can cut carbon emissions between 11 and 22 percent. To do this, we used deep reinforcement learning to optimize one million eco-driving scenarios. 🚗🤖🧠
We found that controlling vehicle speeds to mitigate traffic across a city can cut carbon emissions between 11 and 22 percent. To do this, we used deep reinforcement learning to optimize one million eco-driving scenarios. 🚗🤖🧠
August 11, 2025 at 6:47 PM
This project was 4 years in the making and it's finally out!
We found that controlling vehicle speeds to mitigate traffic across a city can cut carbon emissions between 11 and 22 percent. To do this, we used deep reinforcement learning to optimize one million eco-driving scenarios. 🚗🤖🧠
We found that controlling vehicle speeds to mitigate traffic across a city can cut carbon emissions between 11 and 22 percent. To do this, we used deep reinforcement learning to optimize one million eco-driving scenarios. 🚗🤖🧠
Reposted by Cathy Wu
A new study led by Prof. Cathy Wu and colleagues reveals that eco-driving measures, such as dynamically adjusting vehicle speeds to reduce stopping and excessive acceleration, can cut carbon emissions between 11 and 22 percent. news.mit.edu/2025/eco-dri...
Eco-driving measures could significantly reduce vehicle emissions
Implementing co-driving techniques, like the use of intelligent speed controls to mitigate congestion at traffic lights, can significantly reduce intersection carbon dioxide emissions without…
news.mit.edu
August 11, 2025 at 2:15 PM
A new study led by Prof. Cathy Wu and colleagues reveals that eco-driving measures, such as dynamically adjusting vehicle speeds to reduce stopping and excessive acceleration, can cut carbon emissions between 11 and 22 percent. news.mit.edu/2025/eco-dri...
We’ve got an awesome workshop tomorrow at RLC on real-world RL! Also, I’m moderating the panel, so let me know if you have questions about applying RL. 🤖🧠🤖
📍 The RL4RS Workshop is happening tomorrow at
University of Alberta, as part of @rl-conference.bsky.social.
Join us for a focused day on real-world applications of Reinforcement Learning.
🗓️ Full Schedule: rl4rs.github.io/RL4RS/schedu...
We hope to see you there.
University of Alberta, as part of @rl-conference.bsky.social.
Join us for a focused day on real-world applications of Reinforcement Learning.
🗓️ Full Schedule: rl4rs.github.io/RL4RS/schedu...
We hope to see you there.
August 4, 2025 at 11:18 PM
We’ve got an awesome workshop tomorrow at RLC on real-world RL! Also, I’m moderating the panel, so let me know if you have questions about applying RL. 🤖🧠🤖
Reposted by Cathy Wu
Propose some socials for RLC! Research topics, affinity groups, niche interests, whatever comes to mind!
rl-conference.cc/call_for_soc...
rl-conference.cc/call_for_soc...
RLC Call for Workshops
rl-conference.cc
June 25, 2025 at 1:26 PM
Propose some socials for RLC! Research topics, affinity groups, niche interests, whatever comes to mind!
rl-conference.cc/call_for_soc...
rl-conference.cc/call_for_soc...
Reposted by Cathy Wu
Trying to make Ai2 so effective that people start seriously referring to us as "American DeepSeek" 🙏 lots of hard work to do.
June 22, 2025 at 11:03 PM
Trying to make Ai2 so effective that people start seriously referring to us as "American DeepSeek" 🙏 lots of hard work to do.
Reposted by Cathy Wu
Hiring a postdoc to scale up and deploy RL-based planning onto some self-driving cars! We'll be building on arxiv.org/abs/2502.03349 and learn what the limits and challenges of RL planning are. Shoot me a message if interested and help spread the word please!
Full posting to come in a bit.
Full posting to come in a bit.
Robust Autonomy Emerges from Self-Play
Self-play has powered breakthroughs in two-player and multi-player games. Here we show that self-play is a surprisingly effective strategy in another domain. We show that robust and naturalistic drivi...
arxiv.org
June 21, 2025 at 5:14 PM
Hiring a postdoc to scale up and deploy RL-based planning onto some self-driving cars! We'll be building on arxiv.org/abs/2502.03349 and learn what the limits and challenges of RL planning are. Shoot me a message if interested and help spread the word please!
Full posting to come in a bit.
Full posting to come in a bit.
This is soooooo fun and so blazingly fast.
Obsidian 1.9.2 (early access) is available with improvements for Bases:
- Formula syntax is more expressive and powerful.
- Added number of results in the current view.
- Table cells with long text expand to show the entire content when selected.
- Operator dropdown for filters is now searchable.
- Formula syntax is more expressive and powerful.
- Added number of results in the current view.
- Table cells with long text expand to show the entire content when selected.
- Operator dropdown for filters is now searchable.
![The new formula syntax is more flexible, easier to use, and better suited to extensibility. For those familiar with Javascript, the new syntax should feel familiar. New functions and types are described in our docs.
Some highlights include:
Functions are now object-oriented. Instead of contains(file.name, "Books"), the formula would be file.name.contains("Books").
Functions can now be chained. e.g. property.split(' ').sort()[0].lower()
Property names are no longer wrapped in backticks (`). Instead, to reference properties with spaces or special characters, the syntax is note["Property Name"]
There is a new type system which provides greater control when writing formulas.
New functions, such as link, date and list for converting a value to a different type.
New file properties: file.path, file.links (a list of all internal links in this file), and file.tags (a list of all tags in this file, including frontmatter).
Some functions have been replaced by comparison operators. For example, dateBefore(date1, date2) is now date1 < date2.
Date modifications are now much simpler. Instead of dateModify(date, string), you can use date + string, for example, date("01/01/2025") + "1 year"
For help migrating existing Bases to the new syntax, take a look at the migration guide.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:u5cgub464m4dfjyjgx5f24y3/bafkreigaalc6p2sx4dlwa4iqejvs7i4c7cmm24cb44co3lmulqsu4xy5oa@jpeg)
June 9, 2025 at 3:42 PM
This is soooooo fun and so blazingly fast.
Feeling like a fish learning about water. Insightful thread
Because science typically has practical aims, the standards of good science need to go *beyond* truth. Just exemplifying those epistemic values at the top of the thread isn't good enough. Science should be "fit-for-purpose." 6/
link.springer.com/10.1007/s131...
link.springer.com/10.1007/s131...
Non-epistemic values and scientific assessment: an adequacy-for-purpose view - European Journal for Philosophy of Science
The literature on values in science struggles with questions about how to describe and manage the role of values in scientific research. We argue that progress can be made by shifting this literature’...
link.springer.com
June 4, 2025 at 8:58 PM
Feeling like a fish learning about water. Insightful thread
I’m incredibly happy to have found metascience bluesky. Science is messy and beautiful, and there is a group of people thinking hard about how and why.
June 4, 2025 at 12:19 PM
I’m incredibly happy to have found metascience bluesky. Science is messy and beautiful, and there is a group of people thinking hard about how and why.
I’m pretty shocked. I was just appointed to a serve on a standing committee in April. TRB has definitely helped me find my footing in the field.
TRB is dissolving all committees. This wasn't unexpected, but they're couching it in a technocratic need to "get back to basics" and ease administrative burden when in fact it's a great big shit sandwich that we're all expected to eat and not make a fuss about.
June 3, 2025 at 3:06 AM
I’m pretty shocked. I was just appointed to a serve on a standing committee in April. TRB has definitely helped me find my footing in the field.
Reposted by Cathy Wu
Organizations and conferences that have the potential to partially fill the void left by TRB (a thread):
June 2, 2025 at 7:13 PM
Organizations and conferences that have the potential to partially fill the void left by TRB (a thread):
Science is a journey not a destination.
Lastly to round up the 🧵: Science works doesn't mean every scientific study produces a fact ready for public consumption. Scientific work is incremental, iterative, nonlinear, slow, imperfect, relentless. Continuous improvement upon what was. Science works not despite but because of many failures.
June 2, 2025 at 12:23 PM
Science is a journey not a destination.
This is the way to cross a freeway. 😍
Breathtaking walkway across SR-520 in Redmond, WA at Microsoft Campus🚶♀️🚲🛣️
Breathtaking walkway across SR-520 in Redmond, WA at Microsoft Campus🚶♀️🚲🛣️
June 2, 2025 at 5:39 AM
This is the way to cross a freeway. 😍
Breathtaking walkway across SR-520 in Redmond, WA at Microsoft Campus🚶♀️🚲🛣️
Breathtaking walkway across SR-520 in Redmond, WA at Microsoft Campus🚶♀️🚲🛣️
Great advice in here. “Aim to have such a high growth rate that you cringe looking back at yourself a year ago.”
Want to get ahead in your early career? 👩💼🚀
Check out Celine Halioua's advice for early career professionals! 🌟📈
Check out Celine Halioua's advice for early career professionals! 🌟📈
Want to get ahead in your early career? 👩💼🚀
Discover tips and strategies for building your skills, networking effectively, and finding your path in your chosen field, and learn from the experience of a successful entrepreneur and business leader.
amplt.de
June 1, 2025 at 12:00 PM
Great advice in here. “Aim to have such a high growth rate that you cringe looking back at yourself a year ago.”
This will mean 75% fewer NSF grants to award…
Looks like they'll achieve that in part by funding about a quarter of the grants they are currently funding. 7% success rate is harrowing stuff. If you're not at a top school, will you even bother applying?
May 31, 2025 at 2:56 AM
This will mean 75% fewer NSF grants to award…
Reposted by Cathy Wu
As citizens concerned about the future #STEM #workforce, we want to hear stories and impacts of #federal #funding cuts on early career #scientists. Take the surveys by June 6! @scienceforgood.bsky.social @jpflores.rbind.io
Impacts survey: bit.ly/3Z1aFPj
Stories survey: bit.ly/3Sghx7R
#savescience
Impacts survey: bit.ly/3Z1aFPj
Stories survey: bit.ly/3Sghx7R
#savescience
May 24, 2025 at 2:51 AM
As citizens concerned about the future #STEM #workforce, we want to hear stories and impacts of #federal #funding cuts on early career #scientists. Take the surveys by June 6! @scienceforgood.bsky.social @jpflores.rbind.io
Impacts survey: bit.ly/3Z1aFPj
Stories survey: bit.ly/3Sghx7R
#savescience
Impacts survey: bit.ly/3Z1aFPj
Stories survey: bit.ly/3Sghx7R
#savescience
📣 I will be giving two talks this week at the TRB DATA conference in Seattle, which explores the intersection of transportation, artificial Intelligence, and data analysis. I love these practitioner + researcher conferences. 🤖🧠👇
May 26, 2025 at 6:02 PM
📣 I will be giving two talks this week at the TRB DATA conference in Seattle, which explores the intersection of transportation, artificial Intelligence, and data analysis. I love these practitioner + researcher conferences. 🤖🧠👇
Reposted by Cathy Wu
📢 Excited to share the first X-Sense Workshop: Ego-Exo Sensing for Smart Mobility @iccv.bsky.social with outstanding speakers!!
x-sense-ego-exo.github.io
Please stay tuned for the Call for Papers
#ICCV2025 #ICCV #ICCV25
#NYCU #Cornell #NVIDIA #USYD #MIT #UCSD #TUDelft #UCLA #NEC #Stanford
x-sense-ego-exo.github.io
Please stay tuned for the Call for Papers
#ICCV2025 #ICCV #ICCV25
#NYCU #Cornell #NVIDIA #USYD #MIT #UCSD #TUDelft #UCLA #NEC #Stanford
May 21, 2025 at 4:38 PM
📣 #RERITE 's first conference presentation will take place tomorrow at the Transportation Research Symposium at Rotterdam in The Netherlands! 🚂🛣️
We use Large Language Models (LLMs) to measure the state of data & code availability in transportation research. Join to learn: 👇
We use Large Language Models (LLMs) to measure the state of data & code availability in transportation research. Join to learn: 👇
May 25, 2025 at 9:35 PM
📣 #RERITE 's first conference presentation will take place tomorrow at the Transportation Research Symposium at Rotterdam in The Netherlands! 🚂🛣️
We use Large Language Models (LLMs) to measure the state of data & code availability in transportation research. Join to learn: 👇
We use Large Language Models (LLMs) to measure the state of data & code availability in transportation research. Join to learn: 👇
Reposted by Cathy Wu
If you’re working on machine learning for combinatorial optimization, a new benchmark just dropped. What makes it different from the ones used in individual papers?
1) It covers 8 standard problems
2) It comes from recognized datasets and competitions
3) It’s really hard!
arxiv.org/abs/2505.16952
1) It covers 8 standard problems
2) It comes from recognized datasets and competitions
3) It’s really hard!
arxiv.org/abs/2505.16952
A Comprehensive Evaluation of Contemporary ML-Based Solvers for Combinatorial Optimization
Machine learning (ML) has demonstrated considerable potential in supporting model design and optimization for combinatorial optimization (CO) problems. However, much of the progress to date has been e...
arxiv.org
May 23, 2025 at 3:14 PM
If you’re working on machine learning for combinatorial optimization, a new benchmark just dropped. What makes it different from the ones used in individual papers?
1) It covers 8 standard problems
2) It comes from recognized datasets and competitions
3) It’s really hard!
arxiv.org/abs/2505.16952
1) It covers 8 standard problems
2) It comes from recognized datasets and competitions
3) It’s really hard!
arxiv.org/abs/2505.16952
Reposted by Cathy Wu
News!⚡
We’re pleased to announce a new Call for Contributed Talks to bring novel and diverse viewpoints to EWRL2025.
Early-career researchers are especially encouraged to submit proposals and share their work with the community.
Full details: euro-workshop-on-reinforcement-learning.github.io/ewrl18/
We’re pleased to announce a new Call for Contributed Talks to bring novel and diverse viewpoints to EWRL2025.
Early-career researchers are especially encouraged to submit proposals and share their work with the community.
Full details: euro-workshop-on-reinforcement-learning.github.io/ewrl18/
May 23, 2025 at 8:04 AM
News!⚡
We’re pleased to announce a new Call for Contributed Talks to bring novel and diverse viewpoints to EWRL2025.
Early-career researchers are especially encouraged to submit proposals and share their work with the community.
Full details: euro-workshop-on-reinforcement-learning.github.io/ewrl18/
We’re pleased to announce a new Call for Contributed Talks to bring novel and diverse viewpoints to EWRL2025.
Early-career researchers are especially encouraged to submit proposals and share their work with the community.
Full details: euro-workshop-on-reinforcement-learning.github.io/ewrl18/
