CheckDistill: Adaptive Checklist Distillation for Efficient Distributional Reward Modeling in Image Editing

Key points

• Direct scalar prediction compresses multidimensional editing preferences, whereas generative evaluators can expose their judgment through explicit reasoning.1
• We introduce CheckDistill, a pointwise reward modeling framework that internalizes instruction-adaptive checklists.1
• We decompose each editing condition into atomic, independently verifiable requirements and construct four-dimensional scoring labels together with structured verification evidence.1
• A reliability-weighted on-policy self-distillation objective adjusts supervision at score tokens according to agreement between the teacher's expected scores and reference labels, while retaining candidate probability weights in a truncated token support.1

Sources (1)

• [1]CheckDistill: Adaptive Checklist Distillation for Efficient Distributional Reward Modeling in Image Editing
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 05:05 PM

  Reinforcement learning for instruction-guided image editing requires reward models that provide reliable, accurate, and efficient feedback.

  Direct scalar prediction compresses multidimensional editing preferences, whereas generative evaluators can expose their judgment through explicit reasoning.

Extractive summary: sentences quoted from the sources.