Tom Silver
@tomssilver
Followers
1,405
Following
242
Media
25
Statuses
211
PhD Student @MIT_LISLab and @MITCoCoSci
Joined October 2011
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
JUNGKOOK
• 708543 Tweets
casilla
• 300158 Tweets
NEVER LET GO IS COMING
• 197434 Tweets
#EleccionesMéxico2024
• 190721 Tweets
Flamengo
• 161594 Tweets
Rojo
• 56145 Tweets
Racing
• 55968 Tweets
Maldives
• 55078 Tweets
Whindersson
• 54021 Tweets
بن نافل
• 40863 Tweets
O Vasco
• 34471 Tweets
THE CLIMB
• 29243 Tweets
Sinner
• 26002 Tweets
Sporting
• 24346 Tweets
RIP Rob
• 21599 Tweets
Allan
• 18654 Tweets
Metz
• 16549 Tweets
Rony
• 14405 Tweets
Beileid
• 12944 Tweets
Oviedo
• 12670 Tweets
David Luiz
• 11029 Tweets
Cebolinha
• 11008 Tweets
Rossi
• 10822 Tweets
Moutet
• 10640 Tweets
Pinned Tweet
I'm thrilled to join Princeton's faculty as an assistant professor in the ECE department starting Fall 2025 🐯
Stay tuned for the launch of my lab. We will develop generally helpful robots that learn and plan 🤖
48
9
468
I defended my PhD
@MITEECS
this week! Thanks to everyone who came out. And thanks especially to
@nishanthkumar23
who not only managed the Zoom, but also got me this amazing gift…
21
3
143
This new preprint () is the culmination of a lot of my work with Rohan Chitnis over the last two years.
4
11
67
New preprint: "Learning Neuro-Symbolic Skills for Bilevel Planning" w/ Ashay Athalye, Josh Tenenbaum, Tomas Lozano-Perez and Leslie Kaelbling.
Check it out if you're interested in hierarchical RL, TAMP, and LfD!
Paper:
Video:
1
4
47
Our paper got an oral at CoRL (rhyme!) We proposed CAMPs, a method for learning to generate abstractions of large planning problems. Stop by at
@corl_conf
this week!
Paper:
Video:
Code:
1
5
43
Very excited to be organizing the Workshop on Learning Effective Abstractions for Planning (LEAP) at
#CoRL2023
with
@shah__naman
@georgiachal
@_ericrosen
@davidpaulius
and Beomjoon Kim! We have fantastic speakers lined up.
Website:
Deadline: September 18
0
13
37
Looking forward to the Foundation Models for Decision Making Workshop at
#NeurIPS2022
!
We'll be presenting preliminary work on using large language models for PDDL planning:
These papers also look very interesting & related:
1
0
30
Congratulations to the twitter-less Rohan Chitnis for defending his thesis today!
0
3
30
Excited to co-organize this CoRL 2022 workshop with Rohan Chitnis,
@GregoryJStein
,
@Yezhou_Yang
, and
@jana_kosecka
! Submissions are now open.
0
7
27
Looking forward to
#AAAI23
! On Tuesday, I'll present work on neuro-symbolic learning for robotic planning at the Bridge Session on AI & Robotics.
I'll show some clips from this 1972 video of Shakey the robot and ask: how much progress have we really made?
2
0
26
New work: "Few-Shot Bayesian Imitation Learning with Logic over Programs"
w/ Kelsey Allen, Alex Lew, Leslie Kaelbling, Josh Tenenbaum
Website:
Short versions to appear at ICLR SPiRL workshop and RLDM (say hello!)
1
3
25
If you're interested in LLMs for planning, check out our new preprint: "Generalized Planning in PDDL Domains with Pretrained Large Language Models"
w/ Soham Dan, Kavitha Srinivas, Josh Tenenbaum, Leslie Kaelbling, Michael Katz (
@MITIBMLab
)
0
5
25
I'm really excited about this "planning to practice" direction. We're able to leave Spot alone for hours and it's much improved when we get back. This is a step towards generalist robots that learn to specialize *during deployment*. Check out the paper!
Can we get robots to improve at long-horizon tasks without supervision?
Our latest work tackles this problem by planning to practice!
Here's a teaser showing initial task -> autonomous practice -> eval (+ interference by a gremlin👿)
1
11
92
0
0
22
Excited to learn that this work will be a spotlight talk at
#RLDM2022
!
This new preprint () is the culmination of a lot of my work with Rohan Chitnis over the last two years.
4
11
67
0
2
21
I'm trying to collect recent benchmarks for generalization in RL. Below are the ones that I've found so far. Please reply if you know of others, or if I'm missing a state of the art.
2
4
21
Noticing recent interest in learning "skill priors" or "behavioral priors":
- SPiRL ()
- Parrot ()
- OPAL ()
Any others? Interesting direction!
3
0
19
Looking forward to AAAI this week! We’ll be presenting two papers on learning for planning in relational domains...
1
4
19
This year, Rohan Chitnis and I have been trying to figure out how we can learn models through environment interaction, like in model-based RL, and then use the models to plan w/ powerful robotic planners, like those found in task and motion planning (TAMP).
1
1
19
In the meantime, I'm also very excited to do a postdoc at Cornell in the
@EmpriseLab
with
@TapoBhat
starting this summer!
0
1
18
I'm very excited to co-organize this workshop! The submission deadline is May 19. Consider submitting your work on learning for TAMP!
We are organizing the RSS’23 Workshop on Learning for Task and Motion Planning
Contributions of short papers or Blue Sky papers are due May 19th, 2023.
2
8
41
0
1
16
cognitive psychology: is X built in?
AI in theory : should X be built in?
AI in practice: what happens if we build in X?
0
1
15
When is it okay to use a 'max' in reporting results for RL? Two examples:
1. Train on multiple random seeds and report only the best seed
2. At training iteration t, report the max over training iterations 1, 2, ..., t.
(1/10)
2
2
14
It would be cool if AI/ML conferences solicited some survey/review papers. I wonder if that's ever been discussed.
0
0
13
(1/2) New work on "Residual Policy Learning" with Kelsey Allen, Josh Tenenbuam & Leslie Kaelbling:
Simple idea: start with a meh policy, learn a residual function to improve it. Does better than deep RL or initial policy alone.
1
1
13
The submission portal for our RSS workshop on learning for TAMP is now open! The deadline is May 19. We're going to have some really exciting speakers. The format is hybrid, so even if you can't make it to Korea, consider submitting!
1
3
13
Curious about task and motion planning? Check out our new survey paper:
(w/ Caelan Reed Garrett, Rohan Chitnis, Rachel Holladay, Beomjoon Kim, Leslie Pack Kaelbling, Tomás Lozano-Pérez)
0
0
13
The submission deadline for our LEAP Workshop at
@corl_conf
2023 has been extended to Sep 30!
Very excited to be organizing the Workshop on Learning Effective Abstractions for Planning (LEAP) at
#CoRL2023
with
@shah__naman
@georgiachal
@_ericrosen
@davidpaulius
and Beomjoon Kim! We have fantastic speakers lined up.
Website:
Deadline: September 18
0
13
37
0
3
12
New blog post:
TLDR; Problem Setting sections should be standard in AI papers!
0
1
12
I'm really excited for our upcoming
#CoRL2023
workshop on learning effective abstractions for planning (LEAP)! The deadline to submit a paper is coming up quickly (Sep 30). We're accepting submissions in both the CoRL or ICRA formats. Hope to see you there!
5 days left!! Consider submitting your work on learning abstractions for decision-making in robotics in our (
@tomssilver
,
@davidpaulius
,
@_ericrosen
,
@GeorgiaChal
, Beomjoon Kim) workshop
#LEAP
at
#CoRL23
.
0
7
17
0
3
11
Our previous work made the crucial assumption that relational state abstractions, in the form of logical predicates, were provided. The main contribution in our new preprint is to learn these predicates from data, without any supervision on their form or number.
1
0
10
Looking forward to presenting our work on learning neuro-symbolic skills for bilevel planning at
@corl_conf
!
Camera-ready now up:
Video:
0
1
10
What paper do you come back to most often for another read?
For me, it is "From Skills to Symbols: Learning Symbolic Representations for Abstract High-Level Planning" by George Konidaris, Leslie Kaelbing, and Tomas Lozano-Perez. ()
1
0
10
Check out our new blog posts!
New blog posts (two for the price of one!)
Rohan Chitnis and
@tomssilver
discuss their recent work on learning to generate abstractions for faster planning.
1
1
26
0
0
10
Excited to share this new blog post led by
@nishanthkumar23
@williebeit
and Kathryn Le! It's a great primer on bilevel planning, the bedrock of our learning and planning work over the last few years.
Ever heard about "Bilevel Planning" or "Task and Motion Planning", but been unsure what those words mean? Ever wanted a gentle intro to these methods so you can just understand what's going on? Our new blog post might help!
3
18
81
1
0
10
We're getting some great submissions to this CoRL workshop! Add yours by November 1 :)
Excited to co-organize this CoRL 2022 workshop with Rohan Chitnis,
@GregoryJStein
,
@Yezhou_Yang
, and
@jana_kosecka
! Submissions are now open.
0
7
27
0
0
9
“Hi Tom -
This is a bit of a weird one. It turns out that in Ubuntu 22.04 there's now a system user, related to the Trusted Platform Module, that is also called 'tss', and that's breaking your ability to log in.”
Time to legally change my name…
2
0
9
Next (), we considered neuro-symbolic techniques for learning two additional components: (1) samplers, which stochastically refine the abstract actions into low-level controllers; and (2) a low-level transition model.
1
0
8
We finally came up with an objective for learning predicates that is efficient enough to evaluate, but still closely tied to our real planning objective. The main idea is to use demonstration data to cheaply approximate the most expensive parts of bilevel planning.
1
0
8
This is joint work with
@nishanthkumar23
,
@williebeit
, Tomas Lozano-Perez, Leslie Kaelbling, and Josh Tenenbaum. Also, we’re looking forward to presenting an extended abstract at
#RLDM2022
!
0
0
8
I would like to nominate the grading interface on Gradescope for best inventions of the 21st century
0
0
8
Predicate learning proved to be the most challenging part of the pipeline. Among many issues that we encountered, one interesting theme was: state abstractions that are good for making predictions (bisimulation) are not necessarily good for planning.
1
0
8
Engineering AI by Leslie Kaelbling
0
1
7
My lab now has a twitter account!
Hello, world! We are the Learning and Intelligent Systems group
@MIT_CSAIL
, headed by Leslie Pack Kaelbling & Tomás Lozano-Pérez. We work on AI, ML, and robotics, and we’ll be mostly tweeting about new work by our group.
1
10
73
0
0
7
We first focused on learning action abstractions as planning operators () using techniques from inductive logic programming.
1
0
7
Our objective has been to learn all of the models needed for bilevel planning, as in task and motion planning (TAMP). Unlike in model-based reinforcement learning, a low-level transition model is not enough; we also need state and action abstractions for high-level planning.
1
0
7
PLOI: Planning with Learned Object Importance in Large Problem Instances using Graph Neural Networks
Paper:
Code:
Video:
1
3
7
Then on Saturday, I'll present our work on "Predicate Invention for Bilevel Planning" at the main conference (oral at 9:30am in Room 147B; poster at 6:15pm).
Let me know via email if you're at
#AAAI23
and want to chat!
0
0
6
Excellent blog post by
@nishanthkumar23
on one of the central questions in our field right now!
There was a lot of good and interesting debate on "is scaling all we need to solve robotics?" at
#CoRL23
. I spent some time writing up a blog post about all the points I heard on both sides:
22
47
257
0
0
6
In your AI research, which of the following do you wonder most often?
How did evolution find X?
14
How can a child learn X?
20
How did I learn X?
10
Other
4
0
0
6
Also check out
@nishanthkumar23
and
@williebeit
's
#CoRL2023
paper, the latest advance in our effort to learn all the models that you need to do bilevel planning, this time in BEHAVIOR!
Website:
Code:
Paper:
0
0
4
I'm sure this is not original, but "citation types" would be useful for following related work trails. E.g.,
[1] Directly extending approach
[2] Same problem setting ([2b] used as baseline)
[3] Supports an assertion but otherwise tangential
[4] Unrelated self-citation :)
1
0
6
Looking forward to presenting this work on few-shot imitation learning with programmatic policies at AAAI 2020!
Check out the updated arXiv paper here:
0
0
5
looks like another nice benchmark for learning and physical reasoning with tools! see also the “Tools” challenge:
0
0
5
"Taskography" looks like a really cool benchmark hoping to use it in my work!
0
1
5
A nice passage about science and social responsibility (from 1972!)
1
1
4
(2/2) See also (concurrent/independent) work out of Berkeley, Siemens, TUHH:
Very nice results on a real robot.
@jackclarkSF
calls this Franken-RL, which I love.
0
0
4
Looking forward to presenting our work on predicate invention for bilevel planning at
#AAAI23
!
Paper:
Code:
0
0
4
Looking forward to talking about PDDLGym () at the ICAPS PRL workshop today and tomorrow, with Rohan Chitnis. Registration is free -- stop by!
0
0
4
Starting momentarily on the 2nd floor of the Starling (Hub 3)!
Consider stopping by our workshop on learning abstractions for long-horizon sequential decision-making at
#CoRL2023
. Room Hub 3
For more details:
0
1
6
0
0
4
I think this is some progress, but there is still a lot to figure out, especially:
- How can we learn the symbolic predicates? (c.f. work by George Kondaris; Masataro Asai)
- Can we combine with work on learning behavior priors? (see discussion in NSRT paper)
More to come!
0
0
4
Sonic the Hedgehog
Paper: Nichol et al. 2018 ()
Code:
SoTA: PPO with tweaks ()
1
0
4
Very cool and inspiring work by
@KelseyRAllen
,
@realkevinsmith
and Josh Tenenbaum
Our “Tools” challenge is finally out! For our
#CogSci2019
paper (with
@realkevinsmith
, arxiv ), we present a fun game to investigate rapid physical trial-and-error learning in humans and machines.
2
25
88
0
0
3
Wow, the interactive visualizations in this article are really cool!
Very excited to share our interactive article:
A visual introduction to Gaussian Belief Propagation!
It's part proposition paper, part tutorial with interactive figures throughout to give intuition.
Article:
Work with:
@talfanevans
,
@AjdDavison
1/n
7
129
575
0
0
3
CoinRun
Paper: Cobbe et al. ICML 2019 ()
Code:
SoTA: IMPALA
1
0
3
Looking forward to
#RLDM2022
! The last RLDM still ranks at the top of my list of favorite conferences. Email me if you're there & want to chat!
0
0
3
arXiv link now available:
1
0
3
@karpathy
enjoyed your YConf slides! Was wondering - where is this (artificial life?) environment from?
0
0
2
Obstacle Tower
Paper: Juliani et al. AAAI-WS 2019 ()
Code:
SoTA: PPO/Rainbow
1
0
2
First we looked at the case where a low-level physics simulator is known and some discrete predicates are defined. We showed how to learn symbolic operators that can be used with a "search-then-sample" TAMP planner.
Paper:
Video:
1
0
2
@BarackObama
How many more shootings will there have to be before you push for real gun control legislation?
#ESB
0
1
2
Cool work!
RL agents explore randomly. Humans explore by trying potential good behaviors, because we have a prior on what might be useful. Can robots get such behavioral priors? That's the idea in Parrot.
arxiv
web
vid
1
22
128
0
0
1
Next we looked at the harder case where we don't have a known simulator. This led us to Neuro-Symbolic Relational Transition (NSRT) models, which can be learned from transition data and used for bilevel TAMP.
Paper:
Video:
1
0
2
@GoAstroMo
i don't know how that pubmed paper was published without addressing the monster attack hypothesis
1
0
2
GLIB: Efficient Exploration for Relational Model-Based Reinforcement Learning via Goal-Literal Babbling
Paper:
Code:
Video:
1
1
2
1
0
2
GVG-AI + Procedural Generation
Paper: Justesen et al. NeurIPS-WS 2018 ()
Code:
SoTA: A2C with progressive procedural content generation
1
1
1
@mark_riedl
My understanding is that the subtask graphs are like the recipes, and they learn those from data. Tagging the authors in case they want to clarify :)
@sungryulls
@jaejaywoo
0
0
1
The key idea is that a self-imposed constraint, like "I'm going to stay inside the kitchen", make some parts of a planning problem irrelevant, e.g., the weather outside. So if we can learn a good constraint to impose, we can automatically derive a problem abstraction.
1
0
1
@luttequotidien
i dunno if they have the colbert report in italy but your boi was on last night
http://t.co/KUfoJpTm
0
0
0
0
0
1
@ben_gar
I like the Reeses blast or the Oreo blast.
@bonjoursteffyg
likes cranberry limeades and banana shakes with caramel
1
0
1
@hugo_larochelle
Once interactive datasets are more available, how many current UL approaches will be overkill in light of active feature learning?
0
0
1
Classic control with varying physical parameters
Paper: Packer et al. 2018 ()
Code:
SoTA: PPO/A2C
1
0
1