Greedy Is Enough: Sparse Action Discovery in Agentic LLMs

2026-01-13

AI Summary

This work addresses the challenge of action space explosion in intelligent agents by proposing an efficient method to identify a sparse subset of critical actions. Formulating action discovery as a block-sparse recovery problem, the authors design a greedy algorithm inspired by Orthogonal Matching Pursuit that provably recovers the true action set with high probability under structured sparsity assumptions. Theoretical analysis establishes an information-theoretic lower bound, highlighting the necessity of both sparsity and coverage. Notably, the sample complexity of the proposed approach grows only logarithmically with the total number of actions, and the resulting policy is nearly optimal for unseen latent states. This framework thus offers a theoretically grounded and practically viable solution for action pruning in agents operating in extremely large action spaces.

Technology Category

• Search and Optimization: Learning to Search
• Multiagent Systems: Adversarial Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

Modern agentic systems operate in environments with extremely large action spaces, such as tool-augmented language models with thousands of available APIs or retrieval operations. Despite this scale, empirical evidence suggests that only a small subset of actions meaningfully influences performance in a given deployment. Motivated by this observation, we study a contextual linear reward model in which action relevance is governed by a structured sparsity assumption: only a small number of actions have nonzero effects across latent states. We formulate action discovery as a block-sparse recovery problem and analyze a greedy algorithm inspired by Orthogonal Matching Pursuit. Under standard assumptions on incoherence, signal strength, and action coverage, we prove that the greedy procedure exactly recovers the relevant action set with high probability, using a number of samples that scales polynomially in the sparsity level and latent dimension, and only logarithmically in the total number of actions. We further provide estimation error guarantees for refitted parameters and show that the resulting decision rule is near-optimal for new latent states. Complementing these results, we establish information-theoretic lower bounds demonstrating that sparsity and sufficient coverage are necessary for tractability. Together, our results identify sparse action discovery as a fundamental principle underlying large-action decision-making and provide a theoretical foundation for action pruning in agentic systems.

Problem

• sparse action discovery
• agentic LLMs
• large action spaces
• structured sparsity
• action relevance

Innovation

• sparse action discovery
• greedy algorithm
• block-sparse recovery
• agentic LLMs
• action pruning