Trait-Aware Policy Optimization for Autoregressive Multi-Trait Essay Scoring
This work addresses the underexplored problem of effectively post-training autoregressive models for fine-grained multi-trait essay scoring, which is important for automated writing evaluation.
The paper proposes Trait-Aware Policy Optimization (TAPO), a post-training framework for autoregressive multi-trait essay scoring that decomposes rewards along sample and trait dimensions. Experiments show consistent improvements over supervised fine-tuning and scalar-reward baselines across multiple backbone models.
Multi-trait essay scoring aims to provide fine-grained evaluation of writing quality across multiple dimensions. However, how to effectively post-train autoregressive scoring models remains underexplored. In this paper, we propose Trait-Aware Policy Optimization (TAPO), a post-training framework tailored to autoregressive multi-trait scoring. Our method decomposes rewards along both the sample and trait dimensions, combining global scoring consistency, trait-level accuracy, format validity, and inter-trait dependency preservation. In addition, we enhance supervised fine-tuning with enhanced prompts, allowing the model to internalize trait semantics before preference optimization. Experiments across multiple backbone models show that our method consistently improves multi-trait scoring performance over supervised fine-tuning and scalar-reward optimization baselines, demonstrating the effectiveness and transferability of trait-aware post-training for essay scoring.