An Empirical Study and Theoretical Explanation on Task-Level Model-Merging Collapse

TL;DR

This paper investigates why merging independently fine-tuned language models sometimes fails catastrophically, revealing that representational incompatibility causes collapse and providing a theoretical framework to understand these limits.

Contribution

It identifies task-level merging collapse as a key failure mode, characterizes its causes, and offers a theoretical explanation based on rate-distortion theory.

Findings

Representational incompatibility correlates with merging collapse.

Parameter-space conflict metrics show minimal correlation.

Theoretical bounds on task mergeability are established.

Abstract

Model merging unifies independently fine-tuned LLMs from the same base, enabling reuse and integration of parallel development efforts without retraining. However, in practice we observe that merging does not always succeed: certain combinations of task-specialist models suffer from catastrophic performance degradation after merging. We refer to this failure mode as merging collapse. Intuitively, collapse arises when the learned representations or parameter adjustments for different tasks are fundamentally incompatible, so that merging forces destructive interference rather than synergy. In this paper, we identify and characterize the phenomenon of task-level merging collapse, where certain task combinations consistently trigger huge performance degradation across all merging methods. Through extensive experiments and statistical analysis, we demonstrate that representational incompatibility between tasks is strongly correlated with merging collapse, while parameter-space conflict metrics show minimal correlation, challenging conventional wisdom in model merging literature. We provide a theoretical explanation on this phenomenon through rate-distortion theory with a dimension-dependent bound, establishing fundamental limits on task mergeability regardless of methodology.