Textual Equilibrium Propagation for Deep Compound AI Systems

2026-01-28

AI Summary

This work addresses the challenge of gradient explosion or vanishing in deep composite AI systems, where global text-based feedback suffers from excessively long backpropagation paths, limiting overall performance. To overcome this, the paper introduces equilibrium propagation to textual optimization for the first time, proposing a two-phase local learning mechanism that iteratively refines prompts through free and perturbed phases to align local equilibria with global objectives. By eschewing end-to-end backpropagation, the method employs local LLM critics, proximal prompt editing, and forward-only signal propagation, thereby avoiding signal degradation and high computational costs while enabling efficient coordination of black-box large language models. Experiments demonstrate superior accuracy and efficiency over global approaches such as TextGrad in long-horizon question answering and multi-agent tool-use tasks, with performance gains amplifying as system depth increases.

Technology Category

• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP
• Multiagent Systems: Adversarial Agents

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Large language models (LLMs) are increasingly deployed as part of compound AI systems that coordinate multiple modules (e.g., retrievers, tools, verifiers) over long-horizon workflows. Recent approaches that propagate textual feedback globally (e.g., TextGrad) make it feasible to optimize such pipelines, but we find that performance degrades as system depth grows. In particular, long-horizon agentic workflows exhibit two depth-scaling failure modes: 1) exploding textual gradient, where textual feedback grows exponentially with depth, leading to prohibitively long message and amplifies evaluation biases; and 2) vanishing textual gradient, where limited long-context ability causes models overemphasize partial feedback and compression of lengthy feedback causes downstream messages to lose specificity gradually as they propagate many hops upstream. To mitigate these issues, we introduce Textual Equilibrium Propagation (TEP), a local learning principle inspired by Equilibrium Propagation in energy-based models. TEP includes two phases: 1) a free phase where a local LLM critics iteratively refine prompts until reaching equilibrium (no further improvements are suggested); and 2) a nudged phase which applies proximal prompt edits with bounded modification intensity, using task-level objectives that propagate via forward signaling rather than backward feedback chains. This design supports local prompt optimization followed by controlled adaptation toward global goals without the computational burden and signal degradation of global textual backpropagation. Across long-horizon QA benchmarks and multi-agent tool-use dataset, TEP consistently improves accuracy and efficiency over global propagation methods such as TextGrad. The gains grows with depth, while preserving the practicality of black-box LLM components in deep compound AI system.

Problem

• textual gradient explosion
• textual gradient vanishing
• deep compound AI systems
• long-horizon agentic workflows
• global textual feedback propagation

Innovation

• Textual Equilibrium Propagation
• compound AI systems
• textual gradient
• local learning
• forward signaling