RL teaches models to work longer, but reasoning is dependent on domain-specific post-training. Baseten's Head of Model Training @oneill_c sat down with @dwarkesh_sp to explain horizon generalization and what's next at the frontier. Full episode here:

RLUSD crossed $2B in market cap last week, with close to $1B issued on the XRP Ledger. 📶 Live since December 2024, built for real-world utility and designed for enterprise use cases. $RLUSD is just getting started:

Rly chafing against the quanity and duration of tasks involved in unpacking my apartment. The endurance of it. I just have to...keep doing tasks? Even if I don't want to? For many hours? Is this what it's like to have a real job

RLUSD Holders Can Now Earn Weekly XRP Rewards On Binance Binance is extending its RLUSD campaign with a $800,000 ripple:native reward pool for eligible users. The rewards will be distributed through weekly XRP airdrops to qualifying RLUSD holders. Users must hold at least 0.01 RLUSD in eligible Earn or Margin accounts. They must also record at least $500 in weekly Margin or Futures trading volume. The campaign runs through Aug. 14, with rewards distributed every Friday.

RLHF book is in print, so we threw it a party in Seattle. Being in a room full of people who actually care about open post-training was really cool. Thanks @radixark for making the night happen, @ManningBooks for the book, and everyone who came out!

RL hold-out-1 experiment im curious about to study how model specialization relates to model capacity and data allocation for how many domains to specialize for a common way to get a model to specialize in N domains is Multi-Teacher-On-Policy Distillation train N specialized teachers, do OPD with routing for a set of prompts to teach a model all those abilities for a given domain like data viz -> if we remove X% of the teachers (ie. Don’t specialize on those skills), does our performance on data viz increase? how is this affected by which domains get left out? does it suffer if similar domains are removed but benefit if very different domains are? we still see vertical focused specialized models like GPT-Cyber clearly looks like allocating a lot of data and parameters for a given vertical boosts perf in that vertical more than a general purpose mixture points towards a future where the specialist models always win for high value domains