# Optimizing Differentiable Relaxations of Coreference Evaluation Metrics

## TL;DR

This paper introduces a differentiable relaxation of coreference evaluation metrics, enabling direct gradient-based optimization and improving neural coreference system performance without reinforcement learning.

## Contribution

We propose a novel differentiable relaxation of coreference metrics, allowing direct optimization and outperforming methods relying on reinforcement learning.

## Key findings

- Significant performance gains in neural coreference systems
- Differentiable relaxation enables direct metric optimization
- Outperforms reinforcement learning approaches

## Abstract

Coreference evaluation metrics are hard to optimize directly as they are non-differentiable functions, not easily decomposable into elementary decisions. Consequently, most approaches optimize objectives only indirectly related to the end goal, resulting in suboptimal performance. Instead, we propose a differentiable relaxation that lends itself to gradient-based optimisation, thus bypassing the need for reinforcement learning or heuristic modification of cross-entropy. We show that by modifying the training objective of a competitive neural coreference system, we obtain a substantial gain in performance. This suggests that our approach can be regarded as a viable alternative to using reinforcement learning or more computationally expensive imitation learning.