Ben Geist

Senior Research Eng @ramplabs / physics + math nerd / Kate Bush fan

[Ben Geist @b_geist]

We just released an oss version of PorTAL, including training scripts and links to the paper’s models on Hugging Face: https://t.co/kNihflRU8t Looking for alpha testers and feedback, give it a spin and let us know what you think!

[Ben Geist @b_geist]

@VictorSanh @rikiyatakehi @_reza_nayebi cool stuff!

[Ben Geist @b_geist]

@ksanok10 Boys with binders >>>

[Ben Geist @b_geist]

Bjork Hunter x Roman Synaulia

[Ben Geist @b_geist]

Average bjork beat on homogenic

[apollo @apollophanes]

Ancient Roman music is the most unsettling shit I’ve ever heard in my life

[Ben Geist @b_geist]

@RampLabs argon looking 👀👀

[Ben Geist @b_geist]

@alexosmonov Miller high lifes

[Ben Geist @b_geist]

and on the 8th day god made peanut m&m’s

[Ben Geist @b_geist]

@teortaxesTex Taps sign

[Ramp Labs @RampLabs]

https://t.co/bUR1I3msD6

[Ben Geist @b_geist]

@ksanok10 Girl that’s Shabbat

[Ben Geist @b_geist]

@matt_vine Piping

[Ben Geist @b_geist]

@RLanceMartin check out the work we did on dreaming for latent memory through self study: https://t.co/boODUY865h

[Ramp Labs @RampLabs]

https://t.co/bUR1I3msD6

[Ben Geist @b_geist]

@RyanAmiri__ Ryan we miss u

[Ben Geist @b_geist]

Humans form short-term memories throughout the day and consolidate them into long-term memory while we sleep. Dreaming plays a huge role in this process, replaying details from the day in different contexts and scenarios.

I think LLMs should have a similar process for long-term memory consolidation. In this setting, dreaming is self-study. We create a self play game that identifies gaps in existing knowledge and can generate targeted experiences that strengthen memory in those areas. Please read more about our work on this below

[Ramp Labs @RampLabs]

https://t.co/bUR1I3msD6

[Ben Geist @b_geist]

many are saying this

[Anthropic @AnthropicAI]

Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. https://t.co/uUEN4bNkr9

[Ben Geist @b_geist]

@adibvafa @jure @Stanford @kvignesh1420 @Radii2323 @ShirleyYXWu @nikitamounier @ShayanPardis nice to see this out !

[Adib @adibvafa]

🦍 Introducing Memorilla: memory as a modality for LLMs

Memorilla turns everything a frozen LLM needs to remember into just 16 memory embeddings

⚡ It uses up to 4,212× fewer tokens and is 27× faster than full context inference while more accurate

🎯 Pushes the Pareto frontier over RAG and agentic memory in personalization, long-document and knowledge QA

A thread 🧵:

[Ben Geist @b_geist]

@rahulgs agree

[Ben Geist @b_geist]

@jackiehluo :D thank you!

[Ben Geist @b_geist]

@eglyman lets make ai dream too :D

[Ben Geist @b_geist]

@datachad ah forgot to include this 😅 ICL accuracy is 89.0%

[Ben Geist @b_geist]

@a_levitator @RampLabs Then you’d have to come back to sf