@Charteddaily might there be more this year due to recent electorate boundaries restructuring? i myself became a special vote in wellington after walking to my local voting place to discover i had crossed a over a new invisible electorate line
The swearing tends to happen when tasks drag on. None of the runs under 100 steps contained "fuck," while 48 of the 50 runs that used it lasted at least 200 steps.
ProgramBench had the most swearing, with profanity appearing in 20 of 200 runs.
To be clear: we urgently need a few sponsors to support a bold instagram friendly artwork installation adjacent to registration as well as the above art exhibition in the sold out expo hall where your own branding can be prominently displayed in appreciation. ๐๐ค
NeurIPS Creative Track 2026 is seeking sponsors to help fill high-visibility public areas of the NeurIPS convention center with art. If you or your company would like to help make this happen, please DM me this week!
just some early results - slightly cherry-picked, but things seem to improve with scale which is a good sign. kudos to @wesg52@sofroniewn@Jack_W_Lindsey and team for the refrence implmentation which made it easy to explore! https://t.co/vp05nUPrUr
New Anthropic research: A global workspace in language models.
Of everything happening in your brain right now, only a tiny fraction is consciously accessibleโthoughts you can describe, hold in mind, and reason with.
We found a strikingly similar divide inside Claude.
enjoying the new global workspace anthropic paper - especially using the Jacobian lens to provide readout interpretations of prints shown at last month CVPR
here's the same effect with a different print: on the 48 layer gemma3-12b, at layer 47 you get bird, woodpecker, ornothology but upstream at layer 37 its again just "bird birds birds"
If you are at CVPR drop by - six pieces are on display today and tomorrow just behind the Google booth in the #CVPR2026 art show. I'll be there for daily artist talks on the official schedule or DM if you'd like to meet up and chat.
pleased to be showing new screenprints at @CVPRConf this weekend. these use mechanistic interpretability techniques to highlight features shared across a wide family of vision models
Each print's unique backlighting element is provided by a TIPS model. This shows spatial patch level activations highlighting visually where that model attends to when activating its analogous feature.
@wendlerch@jeremyphoward@aryaman2020@jatin_n0@voooooogel Thanks Chris! Another use of this contrastive synthetic data technique from years past was steering text-to-image generators away from putting text in the image. https://t.co/plOxPsOjyO
if it makes you feel better: i also introduced the idea of generating useful steering vectors from contrastive synthetic data in my 2016 paper - a whole section on augmenting inputs with low pass gaussian filter to derive a steering vector that produces less blurry samples. https://t.co/5eHRe3HZJn
saxophone print + top-150 SigLIP image probe
though mknn model agreement (a la platonic representation hypothesis) is not part of the test time compute process, it climbs naturally as the as the optimization evolves