DPO makes the policy its own reward model and drops RLHF's RL loop
Rafailov and colleagues proved the swap targets the same KL-constrained objective as PPO-based RLHF, and no later work overturns that theorem, though the record after it locates DPO's real cost in offline practice.
Peloton cut its way to its first annual profit
The fiscal 2026 profit is real, but $178 million of cost cuts and a single quarter carried it, in a year when revenue, subscribers, and guidance all fell.
beartype type-checks a whole list by reading one random item of it
A list of bytes where the annotation promised strings slips past mypy at author time and trips beartype's decorator at the call, which names the exact item that broke the hint.
Run the valuation backward and read the growth already priced in
Invert the discounted-cash-flow model and Mastercard's $569 price turns into the growth rate the buyer is paying for, a number that shifts with every assumption behind it.
Record flooding puts Indiana under a federal emergency
The White River set a record 24.9-foot crest at Anderson, surpassing its 1913 high, on a day that also narrowed the childhood vaccine schedule and pulled 1.6 million dozen eggs into the highest-risk recall tier.
Two labs ship vulnerability-finding models on opposite terms
OpenAI gates GPT-5.6-Cyber behind identity checks and legal attestations while Z.ai moves to publish GLM-5.3's weights for anyone to run.
A priest, not Parnell, coined the word ‘boycott’
Captain Boycott's neighbors in County Mayo withdrew their labor, service, and trade so completely in the autumn of 1880 that it cost the government some £10,000 to bring in the harvest they left standing.