Arb

Unfocused Research Organization. A consultancy by @gleech & @CharlesD353 & @mishayagudin

[gavin leech (Non-Reasoning) @gleech]

@geoffreyirving seems you're right (Erdos problems, human vs teaming vs AI): +17% diff, nearly significant at 5%

thanks to @nc_znc

[Suspended Reason @suspendedreason]

Join us as we publish hermits, informants, translator-traitors, forum freaks, discord archives, pirated plays, psychiatry bloggers, groupchats and television recaps

[Arb @ArbResearch]

https://t.co/qyxxu5jCFi

[Arb @ArbResearch]

I'm glad to be with you, Samwise Gamgee

[Arb @ArbResearch]

🫡

[Arb @ArbResearch]

Interesting followup claim from @MichaelHassid and friends

[gavin leech (Non-Reasoning) @gleech]

New paper on a long-shot I've been obsessed with for a year:

How much are AI reasoning gains confounded by expanding the training corpus 10000x? How much LLM performance is down to "local" generalisation (pattern-matching to hard-to-detect semantically equivalent training data)?

[Arb @ArbResearch]

@natalia__coelho @nc_znc Agree but still better than SWE-Bench

[Arb @ArbResearch]

@nc_znc took a look at the purported Ant-OAI cyber gap.

[The confounders are severe enough that this could be totally misleading] but the naive view is: Mythos has a decent lead, but OAI isn't far behind.

https://t.co/qMy1660clp

[Peter Wildeford🇺🇸🚀 @peterwildeford]

My two questions: if GPT-5.5 and Mythos were roughly equal...

1.) where is OpenAI's large pile of vulnerabilities they discovered?

2.) why is the White House kinda ignoring it and letting it be open widely, while meanwhile micromanaging Anthropic's release plans?

[Arb @ArbResearch]

@idavidrein for 200 more such moments of awe, see

https://t.co/iG5v8wLZFj

[gavin leech (Non-Reasoning) @gleech]

breakthroughs of the year

[Arb @ArbResearch]

@CharlesD353 https://t.co/JRauEj91Aq

[Arb @ArbResearch]

@nc_znc took a look at the purported Ant-OAI cyber gap.

[The confounders are severe enough that this could be totally misleading] but the naive view is: Mythos has a decent lead, but OAI isn't far behind.

https://t.co/qMy1660clp

[Arb @ArbResearch]

Thanks to Patrick Garrity for maintaining the Anthropic list

https://t.co/G2wOQJzaaJ

[Arb @ArbResearch]

https://t.co/W9TSkfj7r1

[Arb @ArbResearch]

cool start to a grand programme. featuring our soft contamination paper

[David Duvenaud @DavidDuvenaud]

Announcing Talkie: a new, open-weight historical LLM! We trained and finetuned a 13B model on a newly-curated dataset of only pre-1930 data. Try it below!

with @AlecRad and @status_effects 🧵

[Arb @ArbResearch]

@phantom_opus oh it totally should be CC0, will fix

[Arb @ArbResearch]

You can't buy melatonin in the UK. But in the US, it's hard to avoid: it's an ingredient in gas station soft drinks. How general is this kind of thing?

Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine (+ does much more)

[Arb @ArbResearch]

@ericraum I'd take the under on that.

[Arb @ArbResearch]

@krishnanrohit https://t.co/FaAQRfsjna

[rip van wankle @vt__snowflake]

Impressive project requiring an almost egregious amount of handwork

[Arb @ArbResearch]

You can't buy melatonin in the UK. But in the US, it's hard to avoid: it's an ingredient in gas station soft drinks. How general is this kind of thing?

Our new beige site looks at every case where medical regulators disagree about whether to approve a medicine (+ does much more)

[Arb @ArbResearch]

Thanks for huge amounts of tedious work by @Rian_com_, Paul Crowe, Phil Harrison, and Jacob Slosser.

Huge thanks to Emergent Ventures @tylercowen for funding this work. Grudging thanks to @bswud for bullying me into doing something concrete to improve the world for once

[Arb @ArbResearch]

We did all the data work before Claude Code existed; I hope to trick one of you into doing better than us. Should take a weekend (: