Repairing Language Model Pipelines by Meta Self-Refining Competing Constraints at Runtime

TL;DR

This paper presents Meta Self-Refining, a framework that enhances language model pipelines by dynamically detecting and repairing conflicts between soft constraints at runtime, improving efficiency and output quality.

Contribution

It introduces a novel meta-corrective layer that monitors and repairs competing soft constraints in LM pipelines during inference, a significant advancement over existing static approaches.

Findings

Meta Self-Refining successfully repairs constraint conflicts in LM pipelines.

The framework reduces backtracking loops, improving efficiency.

It guides LMs to produce more consistent outputs under conflicting constraints.

Abstract

Language Model (LM) pipelines can dynamically refine their outputs against programmatic constraints. However, their effectiveness collapses when faced with competing soft constraints, leading to inefficient backtracking loops where satisfying one constraint violates another. We introduce Meta Self-Refining, a framework that equips LM pipelines with a meta-corrective layer to repair these competitions at runtime/inference-time. Our approach monitors the pipeline's execution history to detect oscillatory failures. Upon detection, it invokes a meta-repairer LM that analyzes the holistic state of the backtracking attempts and synthesizes a strategic instruction to balance the competing requirements. This self-repair instruction guides the original LM out of a failing refining loop towards a successful output. Our results show Meta Self-Refining can successfully repair these loops, leading to more efficient LM programs.