Speculative Actions: A Lossless Framework for Faster Agentic Systems

2025-10-05

AI Summary

AI agents suffer from high end-to-end latency in environment interaction due to sequential execution and frequent API calls—e.g., hours for chess tasks—severely hindering training, evaluation, and deployment efficiency. Method: We propose the first lossless speculative execution framework for general-purpose agents, adapting CPU-style speculation to agent systems. It supports multi-step action prediction, top-K sampling, and uncertainty-aware decoding. A lightweight predictor generates candidate action sequences; API calls and verification proceed in parallel, enabling computation–communication overlap. Contribution/Results: Evaluated across diverse environments (games, e-commerce, search), our framework achieves up to 55% action prediction accuracy, significantly reduces end-to-end latency, and preserves task performance without degradation. It establishes a scalable, lossless, system-level acceleration paradigm for low-latency agent deployment.

Technology Category

• Multiagent Systems: Agent Communication
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Despite growing interest in AI agents across industry and academia, their execution in an environment is often slow, hampering training, evaluation, and deployment. For example, a game of chess between two state-of-the-art agents may take hours. A critical bottleneck is that agent behavior unfolds sequentially: each action requires an API call, and these calls can be time-consuming. Inspired by speculative execution in microprocessors and speculative decoding in LLM inference, we propose speculative actions, a lossless framework for general agentic systems that predicts likely actions using faster models, enabling multiple steps to be executed in parallel. We evaluate this framework across three agentic environments: gaming, e-commerce, web search, and a "lossy" extension for an operating systems environment. In all cases, speculative actions achieve substantial accuracy in next-action prediction (up to 55%), translating into significant reductions in end-to-end latency. Moreover, performance can be further improved through stronger guessing models, top-K action prediction, multi-step speculation, and uncertainty-aware optimization, opening a promising path toward deploying low-latency agentic systems in the real world.

Problem

• Addresses slow execution of AI agents in sequential environments
• Reduces latency by predicting actions using faster parallel models
• Enables lossless acceleration across gaming, e-commerce, and web systems

Innovation

• Predicts likely actions using faster models
• Enables multiple steps to be executed in parallel
• Uses lossless framework for general agentic systems