today was contacted by open claw affiliate
repiled =)
06.03.2026 19:43
👍 0
🔁 0
💬 0
📌 0
today was contacted by open claw affiliate
repiled =)
openclaw-senator
idk what is "python harness", but looks useful openreview.net/forum?id=g9r...
no eu no us
canada and apac is all I've got
+1 great journal, please consider volunteering for the role of Action Editor at TMLR!
can you please post the recording afterwards to your yt?
job-boards.greenhouse.io/deepmind/job...
I don't think that I'm necessary. Coding agents one shot what I do for weeks, create thousands of lines of code. No idea if they're actually useful. Even at GSoC I see many students try to cheat their way via AI-assisted code. I don't really see why it might be useful and in what way. This is life.
I suck balls at EFGs still, most things came from muscular memory but not concious understanding. Keep working, keep staring at these pdfs. Solve-solve-solve.
📣 Reinforcement Learning Summer School is returning to Milan in 2026!
Co-organized with @ellisunitmilan.bsky.social & designed for Master's and PhD students on RL theory, multi-agent systems, RL & LLMs, real-world applications...
📍 Milan 🇮🇹
📅 3-12 June
⏰ Apply by 27 March
🔗 https://bit.ly/4b2Plhp
it's quite interesting if this work is related because it solves LCP:
arxiv.org/abs/1611.05011
looks like a parallel line of work to me, hm
I more understood poker with raise example because it contains subgames 😀
I didn't get some of discussion either. But the Sørensen's paper looks relevant to extend the current LP solver for EFGs
it's as you said non-trivial but may yield some interesting results even for maximal lotteries, I guess (could and must involve some additional work).
thinking mode hallucinates too much, so some might even find it useless
or are you done with pure LP and want more gradient-based stuff
a good solver would feed generations, though!
btw @sharky6000.bsky.social as I am still eager to implement that issue that you've closed covering sub-perferct NEs, there are perturbed games as well: www.itu.dk/~trbj/papers...
maybe, there's something (quasi-)perfect for normal games as well? (the notion is different as stated in the pdf)?
I like the mission of this company to create AI-assisted gaming experience: iconicgames.io
certainly, impoosible to do in every genre but why not?
how kind of you to propose that as representatives of the countries who initiated the conflict
so, there're wrong words, what does it even mean? or it doesn't go further than a purely personal preference?
the blog is about an "old" (half a year paper) but contains some good applications
adding subgame perfect equilibrium solver to open spiel
found a blog about LLMs as evolutional operators that's worth attention www.cognizant.com/us/en/ai-lab...
lowkey smart: dollar per goal prize-to-performance, makes sense when you target a specific aspect to improve your own game, where the metric is clear
can you task your cats with maintaining your posting schedule>
nice plants!
I think the other thing is to look into sensititivyt analysis from dynamical perspective: I've recently read a lot about naomi.princeton.edu/wp-content/u...
look for evolution of the spectrum as the key to stability
yes, you are right about that
they do a fancy regualisation of this problem to be exact, I would say
arxiv.org/abs/2501.19266
so must be applicable, for sure...
technically, might be better because linear programming is extremely well-studied and its connections with robust control/optimisation
however, all i have ever seen modelled was like in matlab toolboxes; interesting how we're ready to solve the problems efficiently =)
@sharky6000.bsky.social probably saw that because you were cited but robust lotteries with (semi-defined) linear programming rather than sigmoidal optimisation
I didn't judge the paper, but maybe for you it'd be worthy to give it a look arxiv.org/abs/2602.21297
