When Agents Go Quiet: Output Generation Capacity and Format-Cost Separation for LLM Document Synthesis

TL;DR

This paper introduces a theoretical framework and practical methods to prevent output stalling in LLM agents during large document synthesis, significantly reducing token usage and improving reliability.

Contribution

The authors develop the Output Generation Capacity measure, prove a Format-Cost Separation Theorem, and formalize an adaptive strategy for optimal output generation, validated across multiple models and document types.

Findings

Deferred rendering reduces token usage by 48-72%.

Output stalling is eliminated with the proposed methods.

The framework is implemented as an open-source tool, GEN-PILOT.

Abstract

LLM-powered coding agents suffer from a poorly understood failure mode we term output stalling: the agent silently produces empty responses when attempting to generate large, format-heavy documents. We present a theoretical framework that explains and prevents this failure through three contributions. (1) We introduce Output Generation Capacity (OGC), a formal measure of an agent's effective ability to produce output given its current context state - distinct from and empirically smaller than the raw context window. (2) We prove a Format-Cost Separation Theorem showing that deferred template rendering is always at least as token-efficient as direct generation for any format with overhead multiplier $\mu_f > 1$, and derive tight bounds on the savings. (3) We formalize Adaptive Strategy Selection, a decision framework that maps the ratio of estimated output cost to available OGC into an optimal generation strategy (direct, chunked, or deferred). We validate the theory through controlled experiments across three models (Claude 3.5 Sonnet, GPT-4o, Llama 3.1 70B), four document types, and an ablation study isolating each component's contribution. Deferred rendering reduces LLM generation tokens by 48-72% across all conditions and eliminates output stalling entirely. We instantiate the framework as GEN-PILOT, an open-source MCP server, demonstrating that the theory translates directly into a practical tool.