Alli Minetti
48 posts
Alli Minetti retweetledi
The Human Creativity Benchmark is now:
published on @arxiv
dataset on @huggingface
accepted at @icmlconf genAI workshop
who else is going to ICML ?
English
Alli Minetti retweetledi
To kick off Config, we hosted Creative Intelligence, a night on what happens when AI gives you the unexpected, and why, in creative work, the surprising output is often the one worth keeping. We got into how teams use taste, evaluation, and iteration to tell the difference.
Massive thanks to our panelists @adele__li (Product, ChatGPT Images, @OpenAI), @joanrod_ai (CEO/Founder, @QuiverAI), @rodrigodavies (AI Product Leader, @figma), and @viccpoes (CEO/Founder, @krea_ai) for the great conversation, to @joinaicircle for hosting with us, and to everyone who came out!
Such a fun night. Excited for more conversations like these.
English
Alli Minetti retweetledi
As a designer, I love a crit.
It strengthens and sharpens your eye to take a step back and look at designs from a fresh perspective.
Turns out that works with AI too, getting outputs closer to what humans consider "good".
ben@contraben
We all know AI has a "taste" problem. We measured it. It’s (somewhat) fixable, but AI judges can't replace human creative judgement. Introducing Design Crit. New research from @contralabs_ai and @world_lica, now live on @arxiv
English
So excited this is out!! Hard work from us at @contralabs_ai and @world_lica. A big thank you to the amazing Contra community designers who participated. Give it a read on archive!
ben@contraben
We all know AI has a "taste" problem. We measured it. It’s (somewhat) fixable, but AI judges can't replace human creative judgement. Introducing Design Crit. New research from @contralabs_ai and @world_lica, now live on @arxiv
English
Alli Minetti retweetledi
The Config Makeathon is officially LIVE.
$100K in prizes to bring your ideas to life.
Figma@figma
Design matters now more than ever. The Config Makeathon starts now.
English
I mean... who doesn't want Labs swag? Find us at CVPR, say hey, get some merch!
Allison Nulty (ICML26🇰🇷)@ajnulty
Come find me @CVPR for some @contralabs_ai merch 👀
English
Alli Minetti retweetledi
Alli Minetti retweetledi
For @GeminiApp, your prompt is the difference between Client V1 and Production Ready.
We observed 10 designers going through a real world campaign workflow:
>Hero stills
>Social cuts
>Secondary assets
total: 29 deliverables.
Only 24% were "Production Ready"
And they all wrote prompts the same way.
Below are the prompts and outputs 👇⬇️
Contra Labs@contralabs_ai
English
Design agents on the canvas opens up so many doors for how we design and build products.
The first step is having agents that meet you where you are, and design with you.
I’m so stoked for the countless of figmates who put so much into this.
Over a year training our own models and hillclimbing on quality. Truly a labour of love.
See you in the Beta :)
Figma@figma
wdym of course there’s an agent right on the design canvas that’s fluent in Figma and native to the way your team works
English
Alli Minetti retweetledi
Which image model is actually good for real world creative briefs?
We tested:
>4 frontier image models.
>6 brand campaigns.
>Blind eval by a bespoke panel of expert creatives.
@OpenAI's GPT Image 2 wins overall (60% pairwise win rate, 79 Elo clear of 2nd).
but... when you sort by brief and the leaderboard inverts 👀
Contra Labs@contralabs_ai
English
Alli Minetti retweetledi
Wow, killer work from the Krea team 👀
Krea 2 Large is the #2 style-transfer model, already closing in on GPT Image 2.
Average Style Fidelity gap from @krea_ai's latest model to @OpenAI's GPT Image 2: 0.14 points.
Gap from Krea to the rest of the field: more than 4x larger.
Full study below.
Krea@krea_ai
this is Krea 2. our first foundation model, built completely from scratch for aesthetic diversity and stylistic control. learn more and get early access 👇
English
Check it out! We've got more coming, keep up with our research on Contra Labs - don't miss out!! 👀
Contra Labs@contralabs_ai
When experts disagree about AI work, most benchmarks just average the scores. But when you train a model on average opinions, you get average outputs.
English
Alli Minetti retweetledi
Thoughtful work on the Human Creativity Benchmark from @contralabs_ai
I especially appreciated the convergence / divergence framing against the phases of creative workflows. I think it's important to start talking about AI output with the nuances of time, audience, intent
As an industry, we're underestimating the importance of divergence as a feature in the creative process...not a bug
English
Alli Minetti retweetledi
Alli Minetti retweetledi
We ran a blind head-to-head on the leading image models for one specific job:
>product detail shots.
Seedream 5.0 Lite (@BytePlusGlobal, @BytedanceTalk) beat the flagship models from Google, OpenAI, and Black Forest Labs.
It won 2 out of 3 times.
Contra Labs@contralabs_ai
English