Pith. sign in

Title resolution pending

A similar line of reasoning shows that when ℓ(v) = (1 − v)2, the link function satisfies η = 1+v⋆ 2, so the resulting reward model is P(ω|x, y, y′) = 1+R⋆(x,y)−R⋆(x,y′) 2 · 2017

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
browse 1 citing papers

Title metadata for this work has not finished resolving. The hub is built from the citation graph; the title resolver retries DOI and OpenAlex on its next pass.

fields

cs.LG 1

years

2025 1

verdicts

CONDITIONAL 1

representative citing papers

Design Considerations in Offline Preference-based RL

cs.LG · 2025-02-08 · conditional · novelty 6.0

A unified theory of offline RLHF shows that loss curvature and data coverage control suboptimality, explaining why squared-loss IPO is more stable than logistic-loss DPO.

citing papers explorer

Showing 1 of 1 citing paper.

  • Design Considerations in Offline Preference-based RL cs.LG · 2025-02-08 · conditional · none · ref 3

    A unified theory of offline RLHF shows that loss curvature and data coverage control suboptimality, explaining why squared-loss IPO is more stable than logistic-loss DPO.