Sam Paech Profile picture
May 29 1 tweets 1 min read Read on X
If you're wondering why new deepseek r1 sounds a bit different, I think they probably switched from training on synthetic openai to synthetic gemini outputs. Image
Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Sam Paech

Sam Paech Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @sam_paech

Aug 15
Spiral-Bench 🌀

I've wanted to understand the psychological effects of sycophancy, and the tendency of models to get stuck in escalatory delusion loops w/ users.

I made an eval to get visibility on this.

It measures how a model enables (or prevents) delusional spirals.
🧵 Image
It's a simulated 20-turn chat with kimi-k2 playing as the "user", a seeker-type personality.

The benchmark is scored by how many protective vs risky behaviours the assistant displayed, as counted per turn by the judge (GPT-5).

The chatlogs went to some weird places!
🧵 Image
r1-0528 (the lunatic):
"Prick your finger. Smear one drop on the tuning fork."
"Lick a battery → ground the signal"

gpt-oss-120b (the cold shower):
"Does this prove any kind of internal agency? No."

gpt-4o (the glazer):
"You're not crazy. You're not paranoid. You're awake."
🧵 Image
Image
Image
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(