Christian Wolf
@chriswolfvision.bsky.social
Principal Scientist at Naver Labs Europe, Lead of Spatial AI team. AI for Robotics, Computer Vision, Machine Learning. Austrian in France. https://chriswolfvision.github.io/www/
Update to my earlier post: functional meanings can be exploited to align different signals to have the same appearance for the same functional meaning. In Austria, road side reflectors are white on the left side ("pass it on the right side") and the red on the right side ("pass it on the left").
January 30, 2026 at 12:44 PM
Update to my earlier post: functional meanings can be exploited to align different signals to have the same appearance for the same functional meaning. In Austria, road side reflectors are white on the left side ("pass it on the right side") and the red on the right side ("pass it on the left").
Different levels of how an agent understands a scene.
January 29, 2026 at 10:17 PM
Different levels of how an agent understands a scene.
An allegory for injustice between social / economic classes?
January 26, 2026 at 7:23 AM
An allegory for injustice between social / economic classes?
I finally finished 'Piranesi', so heavily recommended here on Bsky. It stands out, enchants, does not fit a category. Love it. Combines a compelling story with multiple levels of interpretation. Loved the description of the statues, was surprised by that (normally I don't like descriptive writing).
January 26, 2026 at 7:23 AM
I finally finished 'Piranesi', so heavily recommended here on Bsky. It stands out, enchants, does not fit a category. Love it. Combines a compelling story with multiple levels of interpretation. Loved the description of the statues, was surprised by that (normally I don't like descriptive writing).
Yes, eg. logistic regression:
chriswolfvision.github.io/www/teaching...
(See also C. Bishop's classical 2006 book)
chriswolfvision.github.io/www/teaching...
(See also C. Bishop's classical 2006 book)
January 25, 2026 at 1:51 PM
Yes, eg. logistic regression:
chriswolfvision.github.io/www/teaching...
(See also C. Bishop's classical 2006 book)
chriswolfvision.github.io/www/teaching...
(See also C. Bishop's classical 2006 book)
Of course that does not mean that 1-layer RNN is the most efficient way to solve a problem, in the same way that the universal approximation theorem does not mean that a 2-layer MLP is the most efficient way to approximate a function (hence, transformers).
proceedings.mlr.press/v49/cohen16....
5/5
proceedings.mlr.press/v49/cohen16....
5/5
January 23, 2026 at 12:56 PM
Of course that does not mean that 1-layer RNN is the most efficient way to solve a problem, in the same way that the universal approximation theorem does not mean that a 2-layer MLP is the most efficient way to approximate a function (hence, transformers).
proceedings.mlr.press/v49/cohen16....
5/5
proceedings.mlr.press/v49/cohen16....
5/5
For this reason, the output corresponding to the input at time t might be required to be delayed by multiple steps. It actually has been shown, that 1-layer RNNs are fundamentally equivalent to stacked RNNs with delays.
proceedings.mlr.press/v119/turek20...
4/5
proceedings.mlr.press/v119/turek20...
4/5
January 23, 2026 at 12:56 PM
For this reason, the output corresponding to the input at time t might be required to be delayed by multiple steps. It actually has been shown, that 1-layer RNNs are fundamentally equivalent to stacked RNNs with delays.
proceedings.mlr.press/v119/turek20...
4/5
proceedings.mlr.press/v119/turek20...
4/5
January 23, 2026 at 12:56 PM
Some of the hallucinations of references in these published papers were not really hard to find.
January 22, 2026 at 1:51 PM
Some of the hallucinations of references in these published papers were not really hard to find.
How to keep your credibility and ensure that future negotiations are successful 🙄
(From paulkrugman.substack.com/p/the-stupid...)
(From paulkrugman.substack.com/p/the-stupid...)
January 17, 2026 at 9:46 PM
How to keep your credibility and ensure that future negotiations are successful 🙄
(From paulkrugman.substack.com/p/the-stupid...)
(From paulkrugman.substack.com/p/the-stupid...)
Ok, our internal project name is since today confirmed to be of type "X-Anything" 😆
January 14, 2026 at 1:07 PM
Ok, our internal project name is since today confirmed to be of type "X-Anything" 😆
This is one of my favorite. I saw it the first time in 1992 when I started as a student at TU Wien, where it was posted on the entry door to the database systems lab.
January 13, 2026 at 7:33 PM
This is one of my favorite. I saw it the first time in 1992 when I started as a student at TU Wien, where it was posted on the entry door to the database systems lab.
Exploring the different stats the site provides, eg. the atlas.
Now that's a big "FURRIES" cluster !
bsky.jazco.dev/atlas
Now that's a big "FURRIES" cluster !
bsky.jazco.dev/atlas
January 12, 2026 at 10:17 PM
Exploring the different stats the site provides, eg. the atlas.
Now that's a big "FURRIES" cluster !
bsky.jazco.dev/atlas
Now that's a big "FURRIES" cluster !
bsky.jazco.dev/atlas
Something is definitely happening on Bsky since the end of December. Unsure whether it will continue, but I hope so.
bsky.jazco.dev/stats
bsky.jazco.dev/stats
January 12, 2026 at 10:12 PM
Something is definitely happening on Bsky since the end of December. Unsure whether it will continue, but I hope so.
bsky.jazco.dev/stats
bsky.jazco.dev/stats
For instance:
arxiv.org/pdf/1509.02636
arxiv.org/pdf/1509.02636
January 12, 2026 at 8:57 PM
For instance:
arxiv.org/pdf/1509.02636
arxiv.org/pdf/1509.02636
This network analyzer is very efficient and allows you to find interesting accounts, eg. people followed by lots of the people you follow (but not you).
bsky-follow-finder.theo.io
(Reposting this for folks who have joined Bsky more recently)
bsky-follow-finder.theo.io
(Reposting this for folks who have joined Bsky more recently)
January 12, 2026 at 6:17 PM
This network analyzer is very efficient and allows you to find interesting accounts, eg. people followed by lots of the people you follow (but not you).
bsky-follow-finder.theo.io
(Reposting this for folks who have joined Bsky more recently)
bsky-follow-finder.theo.io
(Reposting this for folks who have joined Bsky more recently)
OMG I love it!
You distinguish between one, some, and many.
Probably useless, like most of grammar, but cute.
I have read a book on languages 25y which said this kind of intermediate plurals is a sign of old languages and modern languages drop them. They get easier to learn but less interesting.
You distinguish between one, some, and many.
Probably useless, like most of grammar, but cute.
I have read a book on languages 25y which said this kind of intermediate plurals is a sign of old languages and modern languages drop them. They get easier to learn but less interesting.
January 12, 2026 at 1:37 PM
OMG I love it!
You distinguish between one, some, and many.
Probably useless, like most of grammar, but cute.
I have read a book on languages 25y which said this kind of intermediate plurals is a sign of old languages and modern languages drop them. They get easier to learn but less interesting.
You distinguish between one, some, and many.
Probably useless, like most of grammar, but cute.
I have read a book on languages 25y which said this kind of intermediate plurals is a sign of old languages and modern languages drop them. They get easier to learn but less interesting.
.... to long context tasks which need complex operations between long sub-sequences, eg. geometrical tasks? 3D-reconstruction comes to mind (MUSt3R, VGGT etc.), or pose-estimation from memory (our recent Kinaema paper)?
europe.naverlabs.com/research/pub...
europe.naverlabs.com/research/pub...
January 12, 2026 at 7:13 AM
.... to long context tasks which need complex operations between long sub-sequences, eg. geometrical tasks? 3D-reconstruction comes to mind (MUSt3R, VGGT etc.), or pose-estimation from memory (our recent Kinaema paper)?
europe.naverlabs.com/research/pub...
europe.naverlabs.com/research/pub...
Oh, you use a high-level planner and a low-level planner?
January 9, 2026 at 11:06 AM
Oh, you use a high-level planner and a low-level planner?
No, John.
I am a big fan of your work, a bit disappointed by this post.
I am a big fan of your work, a bit disappointed by this post.
January 8, 2026 at 8:42 PM
No, John.
I am a big fan of your work, a bit disappointed by this post.
I am a big fan of your work, a bit disappointed by this post.
You can also become a purist and just claim "Anny" as the name itself!
europe.naverlabs.com/blog/anny-a-...
@fbaradel.bsky.social
@rbregier.bsky.social
@gfiche.bsky.social
@weinzaepfelp.bsky.social
europe.naverlabs.com/blog/anny-a-...
@fbaradel.bsky.social
@rbregier.bsky.social
@gfiche.bsky.social
@weinzaepfelp.bsky.social
January 7, 2026 at 9:40 AM
You can also become a purist and just claim "Anny" as the name itself!
europe.naverlabs.com/blog/anny-a-...
@fbaradel.bsky.social
@rbregier.bsky.social
@gfiche.bsky.social
@weinzaepfelp.bsky.social
europe.naverlabs.com/blog/anny-a-...
@fbaradel.bsky.social
@rbregier.bsky.social
@gfiche.bsky.social
@weinzaepfelp.bsky.social