PPO

TL;DR for operators An alignment team with a fixed preference dataset faces a deceptively simple decision: train offline with a direct method, or spend more compute to keep generating and evaluating new responses during training. The cheaper route is not always the safer one. The survey by Tarun Raheja and Nilay Pochhi1 highlights a theoretical coverage result under which offline contrastive preference learning needs stronger coverage of possible responses than online reinforcement learning. If useful responses lie outside the regions represented in the fixed dataset, an offline learner has no direct learning signal there. Online methods can generate new data and therefore operate under a weaker, partial-coverage requirement. ...

Training a reasoning model sounds wonderfully modern until the model discovers that “being correct” and “looking correct enough to satisfy the reward” are not the same career path. That is the quiet problem behind reinforcement learning fine-tuning for large language models. The research conversation often treats methods like PPO, GRPO, and DAPO as a sequence of upgrades: first the classic algorithm, then the critic-free group method, then the decoupled-and-dynamically-sampled variant with a nicer acronym. Very tidy. Unfortunately, models do not read product positioning decks. ...

TL;DR for operators The paper is about air-purifying booth placement in Delhi, but the useful business lesson is broader: optimisation is rarely about chasing the loudest metric. In the study, a greedy strategy that targets the highest-AQI cells achieves the highest overall AQI improvement, at 25.76%. The PPO-based strategy is slightly lower on that headline number, at 25.39%, but much stronger on population impact and traffic impact, with zero green-space violations. ...