@papersandnohope.bsky.social
Pinned
I might look smart, however, I am absolutely not.
also workshop "generative AI for finance" recommended by @talkachman.bsky.social on X (formerly known as Twitter)
sites.google.com/view/neurips...
sites.google.com/view/neurips...
December 7, 2025 at 12:22 PM
also workshop "generative AI for finance" recommended by @talkachman.bsky.social on X (formerly known as Twitter)
sites.google.com/view/neurips...
sites.google.com/view/neurips...
yes I like how this paper talks shit about modern MARL methods
December 6, 2025 at 6:40 PM
yes I like how this paper talks shit about modern MARL methods
found a trend: since like 2021-2, the diversity of academic works hasn't been as large as before
people've (except for some) majorly pivoted to a finite set of topics that could be tightly connected with those financed in production
with capacity of venues growing, it doesn't look as good 4 me
people've (except for some) majorly pivoted to a finite set of topics that could be tightly connected with those financed in production
with capacity of venues growing, it doesn't look as good 4 me
December 6, 2025 at 5:52 PM
found a trend: since like 2021-2, the diversity of academic works hasn't been as large as before
people've (except for some) majorly pivoted to a finite set of topics that could be tightly connected with those financed in production
with capacity of venues growing, it doesn't look as good 4 me
people've (except for some) majorly pivoted to a finite set of topics that could be tightly connected with those financed in production
with capacity of venues growing, it doesn't look as good 4 me
Yes, it is how the things should be
A couple years (!) in the making: we’re releasing a new corpus of embodied, collaborative problem solving dialogues. We paid 36 people to play Portal 2’s co-op mode and collected their speech + game recordings.
Paper: arxiv.org/abs/2512.03381
Website: berkeley-nlp.github.io/portal-dialo...
1/n
Paper: arxiv.org/abs/2512.03381
Website: berkeley-nlp.github.io/portal-dialo...
1/n
December 6, 2025 at 9:57 AM
Yes, it is how the things should be
Reposted
Using synthetic data generated by a smaller model and filtering.
December 4, 2025 at 5:01 PM
Using synthetic data generated by a smaller model and filtering.
Reposted
We are 7 hours into today's program, but we are not done yet! In just a few minutes Michael Jordan will be giving today's second #EurIPS keynote on "A Collectivist, Economic Perspective on AI". So load up on coffee and see you there! ☕
December 4, 2025 at 2:50 PM
We are 7 hours into today's program, but we are not done yet! In just a few minutes Michael Jordan will be giving today's second #EurIPS keynote on "A Collectivist, Economic Perspective on AI". So load up on coffee and see you there! ☕
Reposted
BTW @tyrellturing.bsky.social there is a classical result between responding to future predictions (when learning in games) and correlated equilibrium called "calibrated learning". I was hoping to show that predictive RL agents could approximate this.
www.sciencedirect.com:5037/science/arti...
www.sciencedirect.com:5037/science/arti...
www.sciencedirect.com
December 4, 2025 at 3:34 AM
BTW @tyrellturing.bsky.social there is a classical result between responding to future predictions (when learning in games) and correlated equilibrium called "calibrated learning". I was hoping to show that predictive RL agents could approximate this.
www.sciencedirect.com:5037/science/arti...
www.sciencedirect.com:5037/science/arti...
LLM reviews of papers are as bad as AI-generated images there
December 3, 2025 at 8:44 PM
LLM reviews of papers are as bad as AI-generated images there
Reposted
1/ Why does RL struggle with social dilemmas? How can we ensure that AI learns to cooperate rather than compete?
Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.
Preprint🧵👇
(Paper link below.)
Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.
Preprint🧵👇
(Paper link below.)
December 3, 2025 at 7:19 PM
1/ Why does RL struggle with social dilemmas? How can we ensure that AI learns to cooperate rather than compete?
Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.
Preprint🧵👇
(Paper link below.)
Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.
Preprint🧵👇
(Paper link below.)
Reposted
17/ This theory foundation is just the beginning.
The Google Paradigms of Intelligence team is actively working on extensions of this work – expect more to follow!
github.com/paradigms-of...
The Google Paradigms of Intelligence team is actively working on extensions of this work – expect more to follow!
github.com/paradigms-of...
Paradigms of Intelligence Team
Advance our understanding of how intelligence evolves to develop new technologies for the benefit of humanity and other sentient life - Paradigms of Intelligence Team
github.com
December 3, 2025 at 7:19 PM
17/ This theory foundation is just the beginning.
The Google Paradigms of Intelligence team is actively working on extensions of this work – expect more to follow!
github.com/paradigms-of...
The Google Paradigms of Intelligence team is actively working on extensions of this work – expect more to follow!
github.com/paradigms-of...
A methodical and nice algorithm for zero-shot-coordination
"An Efficient End-to-End Training Approach for Zero-Shot Human-AI Coordination"
openreview.net/pdf?id=6ePsu...
"An Efficient End-to-End Training Approach for Zero-Shot Human-AI Coordination"
openreview.net/pdf?id=6ePsu...
openreview.net
December 3, 2025 at 8:13 PM
A methodical and nice algorithm for zero-shot-coordination
"An Efficient End-to-End Training Approach for Zero-Shot Human-AI Coordination"
openreview.net/pdf?id=6ePsu...
"An Efficient End-to-End Training Approach for Zero-Shot Human-AI Coordination"
openreview.net/pdf?id=6ePsu...
love autodiff, more autodiff and stochastic gradient estimators
hope to see something more than julialang, however, — c or python
hope to see something more than julialang, however, — c or python
I'm excited to see whether our idea translates to general MC integration over Jacobians and gradients outside of XAI. Please don't hesitate to talk to us if you have ideas for applications!
December 3, 2025 at 7:46 PM
love autodiff, more autodiff and stochastic gradient estimators
hope to see something more than julialang, however, — c or python
hope to see something more than julialang, however, — c or python
Everyone who sees this post, please reconsider why you have been following me
You're welcome
You're welcome
December 3, 2025 at 1:40 PM
Everyone who sees this post, please reconsider why you have been following me
You're welcome
You're welcome
Reposted
📣We're hiring new research interns for 2026 at MSR Cambridge! If you're interested in ML research (esp. generative AI and/ or decision making agents), please consider applying. It's a great collaborative environment with a very kind and capable team!
apply.careers.microsoft.com/careers/job/...
apply.careers.microsoft.com/careers/job/...
Research Intern - Machine Learning - People Centric AI | Microsoft Careers
In collaboration with your mentor and a diverse team, contribute to solving an ambitious research challenge and translate your results into actionable insights that are relevant to potential applicati...
apply.careers.microsoft.com
December 2, 2025 at 10:02 PM
📣We're hiring new research interns for 2026 at MSR Cambridge! If you're interested in ML research (esp. generative AI and/ or decision making agents), please consider applying. It's a great collaborative environment with a very kind and capable team!
apply.careers.microsoft.com/careers/job/...
apply.careers.microsoft.com/careers/job/...
Reposted
With more equations than usual, I explain how policy gradient gives you a framework to randomly search for random search heuristics.
Random Search for Random Search
Digging into specific applications of policy gradient
www.argmin.net
December 2, 2025 at 3:30 PM
With more equations than usual, I explain how policy gradient gives you a framework to randomly search for random search heuristics.
Reposted
Swadesh Sistla, Max Kleiman-Weiner
Evaluating LLMs in Open-Source Games
https://arxiv.org/abs/2512.00371
Evaluating LLMs in Open-Source Games
https://arxiv.org/abs/2512.00371
December 2, 2025 at 6:26 AM
Swadesh Sistla, Max Kleiman-Weiner
Evaluating LLMs in Open-Source Games
https://arxiv.org/abs/2512.00371
Evaluating LLMs in Open-Source Games
https://arxiv.org/abs/2512.00371
fascinating product by some good people from Edinburgh
trevormcinroe.github.io/terra_nova
@talkachman.bsky.social might've been secretly enjoying it
trevormcinroe.github.io/terra_nova
@talkachman.bsky.social might've been secretly enjoying it
Terra Nova
trevormcinroe.github.io
December 1, 2025 at 8:49 PM
fascinating product by some good people from Edinburgh
trevormcinroe.github.io/terra_nova
@talkachman.bsky.social might've been secretly enjoying it
trevormcinroe.github.io/terra_nova
@talkachman.bsky.social might've been secretly enjoying it
Reposted
The Center for Digital Play at the IT University of Copenhagen is organizing a new conference: Playing Futures. The conference will take place from May 20th to 22nd at the IT University of Copenhagen. You can read more about it here: playingfutures.today
Playing Futures
Playing Futures is a conference about the role of games and play during and after the impending climate and energy catastrophes. The goal of the conference is to serve as a meeting point for the commu...
playingfutures.today
December 1, 2025 at 1:00 PM
The Center for Digital Play at the IT University of Copenhagen is organizing a new conference: Playing Futures. The conference will take place from May 20th to 22nd at the IT University of Copenhagen. You can read more about it here: playingfutures.today
we all are kinda retarded, and it is kind of undisputable
December 1, 2025 at 2:14 PM
we all are kinda retarded, and it is kind of undisputable
Reposted
Alexander Heckett, Vincent Conitzer
Designing Rules for Choosing a Winner in a Debate
https://arxiv.org/abs/2511.23454
Designing Rules for Choosing a Winner in a Debate
https://arxiv.org/abs/2511.23454
December 1, 2025 at 5:38 AM
Alexander Heckett, Vincent Conitzer
Designing Rules for Choosing a Winner in a Debate
https://arxiv.org/abs/2511.23454
Designing Rules for Choosing a Winner in a Debate
https://arxiv.org/abs/2511.23454
this week might be THE week
November 30, 2025 at 8:49 PM
this week might be THE week
Reposted
This is the related work section. I flagged two problems: "Mandelbrot and Mandelbrot (1982)" was written by one Mandelbrot, and it didn't first introduce procedural terrain generation. I thought they'd at least fix the broken ref, but they fixed nothing. ojs.aaai.org/index.php/AA...
November 30, 2025 at 7:17 PM
This is the related work section. I flagged two problems: "Mandelbrot and Mandelbrot (1982)" was written by one Mandelbrot, and it didn't first introduce procedural terrain generation. I thought they'd at least fix the broken ref, but they fixed nothing. ojs.aaai.org/index.php/AA...
There is a nice discussion about definition of equilibrium in economics happening on twitter. Snapshot of an opinion
davidandolfatto.substack.com/p/on-equilib...
davidandolfatto.substack.com/p/on-equilib...
On "Equilibrium" in Economics
Repost: DSGE Theory (June 2016)
davidandolfatto.substack.com
November 30, 2025 at 3:01 PM
There is a nice discussion about definition of equilibrium in economics happening on twitter. Snapshot of an opinion
davidandolfatto.substack.com/p/on-equilib...
davidandolfatto.substack.com/p/on-equilib...