Diego

[Pau Rodríguez @prlz77]

🚀 Excited to share LinEAS, our new activation steering method accepted at NeurIPS 2025! It approximates optimal transport maps e2e to precisely guide 🧭 activations achieving finer control 🎚️ with ✨ less than 32 ✨ prompts!

💻https://t.co/IdZOpwtFXC
📄https://t.co/sfPHk5sT2B

[Pau Rodríguez @prlz77]

Thrilled to share the latest work from our team at @Apple where we achieve interpretable and fine-grained control of LLMs and Diffusion models via Activation Transport 🔥

📄 https://t.co/TYlwxarrWx
🛠️ https://t.co/gciUcwRNqd

1/9 🧵

[Alexandre Lacoste @alex_lacoste_]

Most of our team is at #ICML2024 , reach out if you want to meet.

We'll be presenting WorkArena and BrowserGym:
Poster Session 2 on Tuesday, Hall C 4-9 #610

https://t.co/Q2NMlxeZsu

[ElevenLabs @ElevenLabs]

Here’s an early preview of ElevenLabs Music.

All of the songs in this thread were generated from a single text prompt with no edits.

Title: It Started to Sing

Style: “Pop pop-rock, country, top charts song.”

[Diego @dvd_42]

@alex_lacoste_ @Satellogic This would not exist without your invaluable feedback and expert guidance through every step of the way, so kudos goes to you really<3

[Alexandre Lacoste @alex_lacoste_]

Very happy to see this dataset out. We've been working on this for a while. Please train some awesome models on it!

(sentinel 2 is still uploading, and a paper is in the writing)

Thanks, @Satellogic and @dvd_42 for this immense help.

[Satellogic @Satellogic]

Access open Satellogic high-resolution satellite imagery to test and train your #AI models.

We're excited to announce the release of an open dataset with more than 6 million Satellogic images to help advance the development and use of foundation models for geospatial intelligence. The dataset is available on HuggingFace, an open-source #machinelearning and data science platform.

➡️Read more:
https://t.co/gH8VVcINEs

[Ethan Mollick @emollick]

|￣￣￣￣￣ ￣|
| This can |
| hack AI |
| now. |
| ＿＿＿＿＿_ |
(\__/) ||
(•ㅅ•) ||
/ 　 づ

Paper showing that ASCII art can get around AI guardrails. Its the return of 1980s hackers. https://t.co/1KGozsE4eQ

[Diego @dvd_42]

@oscmansan @eaclmeeting @prlz77 @Saba_A96 @aidanematzadeh @yashgoyal_ @aagrawalAA Congrats!!!

[Oscar Mañas @oscmansan]

I'm happy to share our work MAPL🍁 has been accepted to #EACL2023 @eaclmeeting (main track) 🎉

Shout-out to my wonderful co-authors @prlz77, @Saba_A96, @aidanematzadeh, @yashgoyal_ and @aagrawalAA!

See you in Dubrovnik 👋

[Oscar Mañas @oscmansan]

🔥 Exciting news! The code and pretrained model weights for our #EACL2023 paper MAPL🍁 are now available on GitHub 🎉

https://t.co/yiVq1jivFg

Catch me at the conference next week to learn more about our work or just chat about multimodal vision-language 👁️💬 modeling!

[Diego @dvd_42]

8/8 We encourage you to try your own counterfactuals explainers on our benchmark! Our method paves the way for more informative and accurate evaluations of counterfactual explanation methods in the image domain. See the project at 🔗: https://t.co/DMmUkUzQcm

[Diego @dvd_42]

1/8 Why did the counterfactual cross the road? To get a fair evaluation on our benchmark! 🐓🛣️. I am super excited to announce our new paper on fair evaluation of counterfactual explainability methods! Published at TMLR. 🧵👇

[Diego @dvd_42]

7/8 We found that explainers often produce uninformative counterfactuals by changing causal attributes, most explanations are redundant, and there has been little improvement in recent years. Our evaluation method highlights areas for future research.

[👩‍💻 Paige Bailey @DynamicWebPaige]

"Write a rap battle about PyTorch and JAX." 🤣