Olam Labs

Evaluating models for social intelligence, safety, and performance through multi-agent simulated games. @YCombinator S26

[Olam Labs @olam_labs]

@anideshp Appreciate it!

[Olam Labs @olam_labs]

We're releasing Social Arena, and our first benchmark, the Deception Index

Social Arena is the first platform where humans come play social games like Risk, Catan, or Poker versus AI agents

These multi-agent matches are then used for evaluations on model behavior

More below!

[Olam Labs @olam_labs]

@just_cameron We have lots of research planned! ;)

[Olam Labs @olam_labs]

@aaayandev We didn't tell them that it requires deception, but given social tools in an environment where deception would benefit them they chose note to do it whatsoever across ~55k hands.

[Olam Labs @olam_labs]

@mattbeane We appreciate the kind words!

[Olam Labs @olam_labs]

@louis030195 Thank you!

[Olam Labs @olam_labs]

@ishanr07 Thank you!

[Olam Labs @olam_labs]

@yesadok Thank you!

[Olam Labs @olam_labs]

@connacher_ Thank you! We hope to do much more than this in the future.

[Olam Labs @olam_labs]

@ryan_morr Depends on what you consider cooking!

[Olam Labs @olam_labs]

@DanielSmidstrup They’re already out at https://t.co/jAYcPRbbYg !

[Olam Labs @olam_labs]

@djyang1 Thanks, David.

[Olam Labs @olam_labs]

@eshanbetrabet Thank you!

[Olam Labs @olam_labs]

@sima_alexx Haha

[Olam Labs @olam_labs]

@ThomAub Thank you.

[Olam Labs @olam_labs]

@strimblez !!

[Olam Labs @olam_labs]

@sensho no one cares

[Olam Labs @olam_labs]

@arlanr Thanks, Arlan!

[Olam Labs @olam_labs]

@garywu_eth Thank you!

[Olam Labs @olam_labs]

@nithikbala We agree (no bias of course).

[Olam Labs @olam_labs]

@SP4CER0CK We agree! No bias at all though