AgentTTS: Large Language Model Agent for Test-time Compute-optimal Scaling Strategy in Complex Tasks

2025/07/26 by Fali Wang, Hui Liu, Wang, Fali +19 · 7 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #I.2.7 #Machine Learning (cs.LG) #Natural Language Processing Techniques #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2508.00890

openalex publication_date 2025/07/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Test-time scaling (TTS) enhances the performance of large language models (LLMs) by allocating additional compute resources during inference. However, existing research primarily investigates TTS in single-stage tasks; while many real-world problems are multi-stage complex tasks, composed of a sequence of heterogeneous subtasks with each subtask requires LLM of specific capability. Therefore, we study a novel problem: the test-time compute-optimal scaling in multi-stage complex tasks, aiming to select suitable models and allocate budgets per subtask to maximize overall performance. TTS in multi-stage tasks introduces two fundamental challenges: (i) The combinatorial search space of model and budget allocations, combined with the high cost of inference, makes brute-force search impractical. (ii) The optimal model and budget allocations across subtasks are interdependent, increasing the complexity of the compute-optimal search. To address this gap, we conduct extensive pilot experiments on four tasks across six datasets, deriving three empirical insights characterizing the behavior of LLMs in multi-stage complex tasks. Informed by these insights, we propose AgentTTS, an LLM-agent-based framework that autonomously searches for compute-optimal allocations through iterative feedback-driven interactions with the execution environment. Experimental results demonstrate that AgentTTS significantly outperforms traditional and other LLM-based baselines in search efficiency, and shows improved robustness to varying training set sizes and enhanced interpretability.

Citations

• A Survey on Collaborating Small and Large Language Models for Performance, Cost-effectiveness, Cloud-edge Privacy, and Trustworthiness
• Dynamic Early Exit in Reasoning Models
• Are Retrials All You Need? Enhancing Large Language Model Reasoning Without Verbalized Feedback
• Two Heads are Better Than One: Test-time Scaling of Multi-agent Collaborative Reasoning
• LLMs Can Generate a Better Answer by Aggregating Their Own Responses
• Multi2: Multi-Agent Test-Time Scalable Framework for Multi-Document Processing
• How Far are LLMs from Real Search? A Comprehensive Study on Efficiency, Completeness, and Inherent Capabilities
• Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?
• Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling
• s1: Simple test-time scaling
• Qwen2.5 Technical Report
• The Limits of Inference Scaling Through Resampling
• Large Language Models for Constructing and Optimizing Machine Learning Workflows: A Survey
• A Comprehensive Survey of Small Language Models in the Era of Large Language Models: Techniques, Enhancements, Applications, Collaboration with LLMs, and Trustworthiness
• Fast Best-of-N Decoding via Speculative Rejection
• Catastrophic Failure of LLM Unlearning via Quantization
• Inference Scaling for Long-Context Retrieval Augmented Generation
• Archon: An Architecture Search Framework for Inference-Time Techniques
• Unlocking Memorization in Large Language Models with Dynamic Soft Prompting
• Relation Also Knows: Rethinking the Recall and Editing of Factual Associations in Auto-Regressive Transformer Language Models
• Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models
• AutoM3L: An Automated Multimodal Machine Learning Framework with Large Language Models
• The Llama 3 Herd of Models
• Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
• BoNBoN Alignment for Large Language Models and the Sweetness of Best-of-n Sampling
• AlphaMath Almost Zero: Process Supervision without Process
• AutoMMLab: Automatically Generating Deployable Models from Language Instructions for Computer Vision Tasks
• InfuserKI: Enhancing Large Language Models with Knowledge Graphs via Infuser-Guided Knowledge Integration
• Offline Training of Language Model Agents with Functions as Learnable Weights
• Large Language Models to Enhance Bayesian Optimization
• Large Language Model Agent for Hyper-Parameter Optimization
• TaskBench: Benchmarking Large Language Models for Task Automation
• OVM, Outcome-supervised Value Models for Planning in Mathematical Reasoning
• Alphazero-like Tree-Search can Guide Large Language Model Decoding and Training
• ChatDev: Communicative Agents for Software Development
• LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion
• Reasoning with Language Model is Planning with World Model
• CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing
• Large Language Models for Automated Data Science: Introducing CAAFE for Context-Aware Automated Feature Engineering
• AutoML-GPT: Automatic Machine Learning with GPT
• Self-Evaluation Guided Beam Search for Reasoning
• MLCopilot: Unleashing the Power of Large Language Models in Solving Machine Learning Tasks
• Can GPT-4 Perform Neural Architecture Search?
• Self-Refine: Iterative Refinement with Self-Feedback
• GPT-4 Technical Report
• Planning with Large Language Models for Code Generation
• EvoPrompting: Language Models for Code-Level Neural Architecture Search
• Self-Consistency Improves Chain of Thought Reasoning in Language Models
• BNAI, NO-TOKEN, and MIND-UNITY: Pillars of a Systemic Revolution in Artificial Intelligence
• Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
• The Web as a Knowledge-base for Answering Complex Questions
• Taking the Human Out of the Loop: A Review of Bayesian Optimization

Cited by

• Scaling Unverifiable Rewards: A Case Study on Visual Insights
• The Art of Scaling Test-Time Compute for Large Language Models
• Budget-Aware Tool-Use Enables Effective Agent Scaling
• Unlocking the Power of Multi-Agent LLM for Reasoning: From Lazy Agents to Deliberation
• Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph
• A Survey on Collaborating Small and Large Language Models for Performance, Cost-effectiveness, Cloud-edge Privacy, and Trustworthiness
• Budget-Aware LLM Discovery via Cost-Calibrated Frontier Utility

Related

• Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph
• Every Rollout Counts: Optimal Resource Allocation for Efficient Test-Time Scaling
• CTTS: Collective Test-Time Scaling
• The Art of Scaling Test-Time Compute for Large Language Models
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters