@quokkka.bsky.social
New embedded (inference) agency paper from that one team at google arxiv.org/abs/2608.03958
A game theory for foundation models shows new paths to rational cooperation through similarity inference
As autonomous agents powered by foundation models are increasingly integrated into social and economic systems, understanding the principles governing their collective behavior is essential for ensuri...
arxiv.org
CADO: From Imitation to Cost Minimization for Heatmap-based Solvers in Combinatorial Optimization Hyungseok Song, Deunsol Yoon, Kanghoon Lee, Han-Seul Jeong, Soonyoung Lee, Woohyung Lim Action editor: Aleksandra Faust https://openreview.net/forum?id=fvxx5FOED6 #optimize #reinforcement
i tried to post my first blogging attempts on this platform and no surprise they've got absolutely 0 frictions so what
Lina Alkarmi, Po-Yen Chen, Mingyan Liu Understanding Federated Learning Through the Lens of Mechanism Design: The Role of Data Heterogeneity https://arxiv.org/abs/2608.00364
I'm happy to announce the #NeurIPS2026 workshop TS-LIMITS (Generalization for Time Series in Tight Settings: Latency, Inference, Memory, prIvacy and susTainability), which will be hosted in Paris, France 🇫🇷, Dec 12–13 2026! 🔗 Check out details and call for papers: ts-limits.github.io
Another fantastic paper from Valeo folks, 35M mid-simulator kilometers to develop safe behavior under partial observability valeoai.github.io/Pictura/
Pictura: Perspective-View Self-Play at Scale for Driving
Self-play driving policies trained directly from rendered perspective images, without privileged vectorized observation of the surroundings.
valeoai.github.io
Playing games with my biological knowlegde (no-op) and huge action spaces alexunderch.github.io/posts/antibo...
Viral adaptation is all we need? (Part 1)
Eye-opening problem of antibody synthesis.
alexunderch.github.io
I love my macbook but I think I should run my experiments somewhere else
Preprint: arxiv.org/abs/2510.181... (the paper being in-press at TMLR) OpenReview: openreview.net/forum?id=yIA...
NashPG: A Policy Gradient Method with Iteratively Refined Regularization for Finding Nash Equilibria
Finding Nash equilibria in two-player zero-sum imperfect-information games remains a central challenge in multi-agent reinforcement learning. Recent multi-round regularization methods offer a promisin...
arxiv.org
Me having similar ideas clearly tells the story. Non-stationary perspective is what we need arxiv.org/abs/2607.26358
Post-Training at the Edge of Detectability: A Game-Theoretic Approach to Fine-Tuning
Reinforcement learning (RL) fine-tuning is widely used in language model training to improve model performance on a target task while limiting drift from a reference policy. A standard way to balance ...
arxiv.org
Boning Li, Yu Chen, Longbo Huang Correlated Chance Sampling for Monte Carlo Counterfactual Regret Minimization https://arxiv.org/abs/2607.27035
@sharky6000.bsky.social , may I ask somewhat obvious opinion: in practical literature, people (you!) mostly optimise for external regret. However, in many theoretical papers, people analyse swap-regret, which is stricter but harder to compute. But how "much" harder? What is the gap, in your opinion?
I don't want to point out people for their "research taste", but still want to share this Tim Roughgarden's playlist www.youtube.com/playlist?lis...
Computation As a Universal and Fundamental Concept - YouTube
Professor Tim Roughgarden treats computation as a law of nature, tracing its arc from Turing's proof that some problems are unsolvable to the P versus NP que...
youtube.com
It was fun giving this tutorial. If you missed it, check out calibration-tutorial.github.io where all of our materials are available, including slides, hundreds of pages of lecture notes, an interactive demo, and an annotated reading list.
Aaron Roth@aaroth.bsky.social · last mo.On Monday @ncollina.bsky.social @iraglobusharris.bsky.social and I are giving a tutorial at ICML on (multi)calibration and its applications. You can find slides and an annotated bibliography on the website: calibration-tutorial.github.io as well as an interactive demo of online calibration algs.
arxiv.org/abs/2607.17968 — found from economists
Equilibrium Play Without Mutual Knowledge of Rationality
Equilibrium play in two-player zero-sum games is usually justified via epistemic assumptions, such as mutual knowledge of rationality and beliefs, that go far beyond the rationality of the players. We...
arxiv.org
scientific conferences witness their own sunset and can't do much about it
RLC26 schedule is now live! rl-conference.cc/schedule.html Join us Aug 15–18 for Workshops, Keynotes: Marc Bellemare, Sheila McIlraith, Danijar Hafner, Rika Antonova & Balaraman Ravindran, Oral sessions, Posters — & of course Cirque du Soleil🎪! Full talk & poster schedule coming soon —stay tuned!
Incredibly proud of the Sakana AI team. We have developed an orchestration model right here out of Japan that achieves state-of-the-art performance on real-world cybersecurity benchmarks! 🎌 sakana.ai/fugu-cyber-r...
Sakana AI@sakanaai.bsky.social · 2w agoIntroducing Fugu-Cyber: an update to our Fugu orchestration model. It achieves state-of-the-art performance on real-world security benchmarks, matching cyber-focused frontier models like GPT-5.5-Cyber and Mythos Preview. Release Blog: sakana.ai/fugu-cyber-r... 🐡
imagine this but for openreview post.substack.com/p/against-cl...
Against Claudefishing
Building trust in the AI age
post.substack.com
✈️ Save the date! AAMAS 2027 heads to Hanoi, Vietnam 🇻🇳 | May 3–7 The 26th edition of the field's largest conference on autonomous agents & multiagent systems, in one of Asia's most vibrant cities. Follow for CfP, deadlines & program updates. #AAMAS2027 #MultiAgentSystems
Did a small piece on why making a problem harder might open some new solutions alexunderch.github.io/posts/better...
An example where collective intelligence helps
How a corner case allows us to find a place for collective intelligence in RL optimisation.
alexunderch.github.io
New NERDS publication on fairness of human-AI collaboration in candidate recommendation nerds.itu.dk/2026/06/29/n...
it's not been a week since I got back from the conference, and almost all anxiety and not knowing what's next are back hashtag life
publishing a generative art/physics cellular automaton i have been working on for a while that has reached a releasable state: 🌬️♟️Chess Gas 🌬️. I received this idea in a vision in the 2000s and it's nice to see it in the world. I hope you like it. Many parameters and colours. chessgas.neocities.org
