Tomer Ullman
@tomerullman.bsky.social
Associate Professor, Department of Psychology, Harvard University. Computation, cognition, development.
major kudos to the other authors, especially YingQiao Wang and Eric Bigelow!
November 10, 2025 at 1:57 PM
major kudos to the other authors, especially YingQiao Wang and Eric Bigelow!
this work can be useful for for simulation in ML, but I think it raises interesting questions for those of us in cognitive science who are interested in mental simulation: what is a reasonable 'step size' for the simulation? Can it be changed? Do we do sub-rendering as well as sub-simulation?
November 10, 2025 at 1:57 PM
this work can be useful for for simulation in ML, but I think it raises interesting questions for those of us in cognitive science who are interested in mental simulation: what is a reasonable 'step size' for the simulation? Can it be changed? Do we do sub-rendering as well as sub-simulation?
...even the case where it made things worse (fluids) is interesting, with the analysis suggesting an initial error that is then compounded by each sub-step
November 10, 2025 at 1:57 PM
...even the case where it made things worse (fluids) is interesting, with the analysis suggesting an initial error that is then compounded by each sub-step
we examined "Chain of Time" in 2d and 3d domains, including simple motion, fluid pouring, bouncing, and moving under gravity.
Chain of time improved things on most of them, and improved it more the more fine-grain the chop
(but not for everything)
Chain of time improved things on most of them, and improved it more the more fine-grain the chop
(but not for everything)
November 10, 2025 at 1:57 PM
we examined "Chain of Time" in 2d and 3d domains, including simple motion, fluid pouring, bouncing, and moving under gravity.
Chain of time improved things on most of them, and improved it more the more fine-grain the chop
(but not for everything)
Chain of time improved things on most of them, and improved it more the more fine-grain the chop
(but not for everything)
there are obvious parallels here to both human mental simulation and LLM in-context reasoning (including Chain of Thought), which are indeed our inspiration.
November 10, 2025 at 1:57 PM
there are obvious parallels here to both human mental simulation and LLM in-context reasoning (including Chain of Thought), which are indeed our inspiration.
the idea is pretty well captured by this image:
the baseline method is to give a model some physical stimuli as input and ask it 'what will this look like x steps into the future?"
Chain of Time breaks this into smaller steps; simulate, output, use that output as new input, repeat.
the baseline method is to give a model some physical stimuli as input and ask it 'what will this look like x steps into the future?"
Chain of Time breaks this into smaller steps; simulate, output, use that output as new input, repeat.
November 10, 2025 at 1:57 PM
the idea is pretty well captured by this image:
the baseline method is to give a model some physical stimuli as input and ask it 'what will this look like x steps into the future?"
Chain of Time breaks this into smaller steps; simulate, output, use that output as new input, repeat.
the baseline method is to give a model some physical stimuli as input and ask it 'what will this look like x steps into the future?"
Chain of Time breaks this into smaller steps; simulate, output, use that output as new input, repeat.
Title: LLMs show signs of cognitive decline with age (or whatever)
Intro:
[Standard-LLM/CogSci-opening-2025b]
ANYWAY
[actual start of paper]
Intro:
[Standard-LLM/CogSci-opening-2025b]
ANYWAY
[actual start of paper]
November 9, 2025 at 10:33 PM
Title: LLMs show signs of cognitive decline with age (or whatever)
Intro:
[Standard-LLM/CogSci-opening-2025b]
ANYWAY
[actual start of paper]
Intro:
[Standard-LLM/CogSci-opening-2025b]
ANYWAY
[actual start of paper]
I again caution that that advice isn't meant to fit everyone; personal statements ARE personal and used differently (or not at all). I am only addressing a specific thing that I sometimes see students facing, when they think it HAS to be about challenges met and overcome.
November 7, 2025 at 6:23 PM
I again caution that that advice isn't meant to fit everyone; personal statements ARE personal and used differently (or not at all). I am only addressing a specific thing that I sometimes see students facing, when they think it HAS to be about challenges met and overcome.
a difficulty that Readers face in evaluation is that things blur together, and Personal Statements can be effective if they help convey a memorable impression. You can be memorable in many ways, first and foremost by being yourself.
November 7, 2025 at 6:23 PM
a difficulty that Readers face in evaluation is that things blur together, and Personal Statements can be effective if they help convey a memorable impression. You can be memorable in many ways, first and foremost by being yourself.
But wait, in their PS Alex wrote about their love of dance. Oh, right, Alex -- the dancer!
It isn't that being in dance is evaluated per se, nor that the essay on dance is about the challenges one met in dance classes. It is about giving a picture of who you are.
It isn't that being in dance is evaluated per se, nor that the essay on dance is about the challenges one met in dance classes. It is about giving a picture of who you are.
November 7, 2025 at 6:23 PM
But wait, in their PS Alex wrote about their love of dance. Oh, right, Alex -- the dancer!
It isn't that being in dance is evaluated per se, nor that the essay on dance is about the challenges one met in dance classes. It is about giving a picture of who you are.
It isn't that being in dance is evaluated per se, nor that the essay on dance is about the challenges one met in dance classes. It is about giving a picture of who you are.
think of it this way: if a Reader is going through 50 applications, they tend to blur together. The Reader is trying to think back to Alex. Who is Alex...? Were they the person that wanted to figure out the mind to build better AI? Maybe...or maybe that was one of 10 other people who said that?
November 7, 2025 at 6:23 PM
think of it this way: if a Reader is going through 50 applications, they tend to blur together. The Reader is trying to think back to Alex. Who is Alex...? Were they the person that wanted to figure out the mind to build better AI? Maybe...or maybe that was one of 10 other people who said that?
...ppl sometimes think personal statements HAVE to be about 'challenges met and overcome'.
They don't.
It's obviously ok to write about that, but you don't *have* to. The statements are usually about giving a sense of who you are as a person; that can take on multiple forms.
They don't.
It's obviously ok to write about that, but you don't *have* to. The statements are usually about giving a sense of who you are as a person; that can take on multiple forms.
November 7, 2025 at 6:23 PM
...ppl sometimes think personal statements HAVE to be about 'challenges met and overcome'.
They don't.
It's obviously ok to write about that, but you don't *have* to. The statements are usually about giving a sense of who you are as a person; that can take on multiple forms.
They don't.
It's obviously ok to write about that, but you don't *have* to. The statements are usually about giving a sense of who you are as a person; that can take on multiple forms.
PART 4: PERSONAL STATEMENTS
Honestly I wasn't sure whether to even comment on this; personal statements truly are personal and up to you, and different schools have different things in mind so make sure to read the prompt carefully and all that.
BUT.
Honestly I wasn't sure whether to even comment on this; personal statements truly are personal and up to you, and different schools have different things in mind so make sure to read the prompt carefully and all that.
BUT.
November 7, 2025 at 6:23 PM
PART 4: PERSONAL STATEMENTS
Honestly I wasn't sure whether to even comment on this; personal statements truly are personal and up to you, and different schools have different things in mind so make sure to read the prompt carefully and all that.
BUT.
Honestly I wasn't sure whether to even comment on this; personal statements truly are personal and up to you, and different schools have different things in mind so make sure to read the prompt carefully and all that.
BUT.
sure; these are both quick imperfect examples, I just mean to give a general sense:
question: "I would like to understand how the brain implements mental simulations in approximate world models"
long term goal: "my aim is to be a neuroscientist, directing my own research group"
question: "I would like to understand how the brain implements mental simulations in approximate world models"
long term goal: "my aim is to be a neuroscientist, directing my own research group"
November 7, 2025 at 2:10 AM
sure; these are both quick imperfect examples, I just mean to give a general sense:
question: "I would like to understand how the brain implements mental simulations in approximate world models"
long term goal: "my aim is to be a neuroscientist, directing my own research group"
question: "I would like to understand how the brain implements mental simulations in approximate world models"
long term goal: "my aim is to be a neuroscientist, directing my own research group"
2. Argument for 'less than you think': despite what I said above, try to imagine getting a 'yes' from a program. Do you feel 'woo!' or do you feel 'oh! Nice...if I have to, that's good, I guess'? I would consider dropping the places that make you feel the second way.
November 6, 2025 at 8:51 PM
2. Argument for 'less than you think': despite what I said above, try to imagine getting a 'yes' from a program. Do you feel 'woo!' or do you feel 'oh! Nice...if I have to, that's good, I guess'? I would consider dropping the places that make you feel the second way.
even if you have top grades, top letters, research experience, etc, that puts you in the top 10% bucket, that's still 1 in 100, and the program is only taking 10 people. So, your odds are 1 in 10, and at that point it's almost a coin flip. You should be flipping that coin more times.
November 6, 2025 at 8:51 PM
even if you have top grades, top letters, research experience, etc, that puts you in the top 10% bucket, that's still 1 in 100, and the program is only taking 10 people. So, your odds are 1 in 10, and at that point it's almost a coin flip. You should be flipping that coin more times.
PART 3: HOW MANY PROGRAMS SHOULD I APPLY TO?
The annoying, honest answer is 'not too many, not too few'.
1. Argument for 'more than you think': There are MANY people applying top top programs. Suppose you are 1 of a 1000.
The annoying, honest answer is 'not too many, not too few'.
1. Argument for 'more than you think': There are MANY people applying top top programs. Suppose you are 1 of a 1000.
November 6, 2025 at 8:51 PM
PART 3: HOW MANY PROGRAMS SHOULD I APPLY TO?
The annoying, honest answer is 'not too many, not too few'.
1. Argument for 'more than you think': There are MANY people applying top top programs. Suppose you are 1 of a 1000.
The annoying, honest answer is 'not too many, not too few'.
1. Argument for 'more than you think': There are MANY people applying top top programs. Suppose you are 1 of a 1000.
4. Help your writers: tell them the programs you're applying to, your relevant strengths, and specific things/anecdotes/achievements you are proud of. Many profs appreciate the reminder. But do check first if they want this ("in drafting the letter, would it be useful for me to...?")
November 6, 2025 at 8:51 PM
4. Help your writers: tell them the programs you're applying to, your relevant strengths, and specific things/anecdotes/achievements you are proud of. Many profs appreciate the reminder. But do check first if they want this ("in drafting the letter, would it be useful for me to...?")
3. A 'good' rec is not about being warm, but being effective.
An effective rec talks about you in specifics.
Consider:
"I met Alex a few times, they seem smart, I'm sure they'll do really well'.
this is 9/10 in for positivity, and 0 for effect.
Ask ppl who can write you effective letters.
An effective rec talks about you in specifics.
Consider:
"I met Alex a few times, they seem smart, I'm sure they'll do really well'.
this is 9/10 in for positivity, and 0 for effect.
Ask ppl who can write you effective letters.
November 6, 2025 at 8:51 PM
3. A 'good' rec is not about being warm, but being effective.
An effective rec talks about you in specifics.
Consider:
"I met Alex a few times, they seem smart, I'm sure they'll do really well'.
this is 9/10 in for positivity, and 0 for effect.
Ask ppl who can write you effective letters.
An effective rec talks about you in specifics.
Consider:
"I met Alex a few times, they seem smart, I'm sure they'll do really well'.
this is 9/10 in for positivity, and 0 for effect.
Ask ppl who can write you effective letters.