annoyingreposter.bsky.social
@annoyingreposter.bsky.social
WWIII is going much smoother than people expect it to be
January 17, 2026 at 10:58 AM
here
January 17, 2026 at 7:12 AM
What it shows to me
January 16, 2026 at 4:54 PM
"Registeation" doesn't seem to be open yet...
January 16, 2026 at 4:12 PM
they use open_spiel btw
January 15, 2026 at 11:09 PM
I feel it's a sign to get more into the implementations.
I saw a lot of them across some repositories, like: github.com/maxosmith/ps...

Maybe, maybe it'd be good to contribute something with more focus on larger scale (for example, LMs as ABR oracles...). Don't know
GitHub - maxosmith/psro: Policy-Space Response Oracles framework.
Policy-Space Response Oracles framework. Contribute to maxosmith/psro development by creating an account on GitHub.
github.com
January 15, 2026 at 10:37 PM
@sharky6000.bsky.social by the way did you decide not to move on with veo videos? That daft punk video was cool.
January 15, 2026 at 8:56 PM
Knowing that pineapple pizza was invented in Canada...
January 15, 2026 at 6:42 AM
yes, other than some special solutions like ExIt-OOS your paper does look like a clear pioneering for AZ in this regard. At least, it's what Scholar Labs has given me.

Sorry for the noise...
January 14, 2026 at 5:48 PM
I sent him a mail recently. Won't happen
January 14, 2026 at 5:42 PM
i am sorry I meant generative like "bayesian" because I wrongfully thought that the network performed bayesian belief update, which turned down to be not correct
January 14, 2026 at 5:41 PM
please share
January 14, 2026 at 5:32 PM
That's why I asked. However, it uses a generative model to model public belief states. Generative model shouldn't help with general-sum games, but with partial observability — you have psro for that, iirc. Agree, that work was about for special factored beliefs, but I see the same principle applied.
January 14, 2026 at 5:28 PM