Related papers: Scores Are Not Decisions: Cost-Aware Stopping for Tool Acquisition in LLM Agents

Scores Are Not Decisions: Cost-Aware Stopping for Tool Acquisition in LLM Agents

URL: http://arxiv.org/abs/2607.27083v1
Date: Wed, 29 Jul 2026 16:07:37 GMT
Title: Scores Are Not Decisions: Cost-Aware Stopping for Tool Acquisition in LLM Agents
Abstract summary: We develop cost-aware marginal decision-focused stopping (CAM-DF) over ranked tool prefixes.<n>We evaluate on 1,343 tasks across five tool-use domains.<n>Our approach is state-of-the-art under heterogeneous costs and high cost pressure.
Score: 14.413992677258825
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Abstract: As LLM agents increasingly depend on diverse external services such as search engines, databases, and connectors, agent harnesses face a fundamental tool-selection challenge: acquiring too few tools leaves the task under-informed, while too many adds cost, context load, and privacy exposure. Routers and retrievers can rank candidate tools by relevance, but a ranking alone does not determine how many are worth selecting. Existing approaches leave acquisition under heterogeneous costs unaddressed. We formulate this decision as cost-aware marginal decision-focused stopping (CAM-DF) over ranked tool prefixes, with CAM-DF-lite as a compact interpretable variant. We train directly on the offline gap between stopping now and the best continuation: its sign labels the decision, its magnitude weights each error by the payoff at stake. We prove this objective is Bayes-aligned with the stopping target and that score-only rules are suboptimal under heterogeneous costs. We evaluate on 1,343 tasks across five tool-use domains. On $τ$-bench Retail, CAM-DF attains the highest payoff among deployable methods, with gains over a predict-then-threshold baseline across all five ranking sources and two cost regimes. Our approach is state-of-the-art under heterogeneous costs and high cost pressure, with larger gains under weaker rankings. In live execution, CAM-DF exposes the agent to 37\% fewer tools than full access while maintaining comparable task success. The CAM-DF family is a lightweight pre-execution plugin that turns existing tool rankings into lower-cost acquisition decisions without fine-tuning the underlying LLM.

Related papers

GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented Agents [5.041247893188969]
We introduce GAUGE, a reusable offline protocol that measures whether a gate's ranking matches a grounded verifiable reward.<n> satisfaction carries essentially no information about task success.<n>We propose a calibrate-then-trust cadence in which a judge-free completion bit is a zero-cost tripwire for truncation regressions.
arXiv Detail & Related papers (2026-09-10T20:29:28Z)
One Policy, Any Budget: Internalizing Budget-Aware Search via Reinforcement Learning [41.55260785497756]
We propose AnySearch, a framework that enables a single policy to perform budget-aware search under any budget constraint.<n>In the first phase, we train the agent with explicit budget state injection and structured reasoning prompts.<n>In the second phase, the scaffold is removed and the agent learns to operate autonomously under adaptively sampled budget constraints.
arXiv Detail & Related papers (2026-09-01T07:12:32Z)
TAPO: Tool-Aware Policy Optimization via Credit Transfer for Multimodal Search Agents [54.08846865906602]
We characterize credit misassignment as a systematic failure mode of GRPO in tool-augmented multimodal search agents.<n>We propose Tool-Aware Policy Optimization (TAPO), which exploits the parameter-determinism property of information-acquisition tools.
arXiv Detail & Related papers (2026-06-04T07:15:43Z)
ZEBRA: Zero-shot Budgeted Resource Allocation for LLM Orchestration [8.226365534099399]
We propose ZEBRA, a framework that reduces multi-phase budget allocation to a continuous nonlinear knapsack problem.<n>On a $150$-task APPS coding benchmark, both ZEBRA variants outperform LLM-direct on every aggregate metric.<n>We show that lightweight algorithmic guidance at inference time can improve the economic behavior of autonomous multi-agent systems.
arXiv Detail & Related papers (2026-05-19T20:50:05Z)
ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox [61.862814740220806]
$textbfComplexMCP$ is a benchmark designed to evaluate agents in rigorous conditions.<n>Built on the Model Context Protocol (MCP), $textbfComplexMCP$ provides over 300 meticulously tested tools.
arXiv Detail & Related papers (2026-05-11T16:20:51Z)
When Is Rank-1 Steering Cheap? Geometry, Granularity, and Budgeted Search [61.73232505100331]
Activation steering offers a lightweight way to control LLMs without retraining, but its effectiveness varies sharply across concepts.<n>We formalize rank-1 steering as a budget-constrained optimization over intervention layer and coefficient.<n>We present textitGRACE, a Granularity- and Representation-Aware Concept Engineering framework that uses activation geometry to diagnose the dominant source of steering difficulty.
arXiv Detail & Related papers (2026-05-09T14:26:49Z)
AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agent [18.58248061903799]
We introduce AgentOpt, the first framework-agnostic Python package for client-side AI optimization.<n>We first study model selection, a high-impact optimization lever in multi-step agent pipelines.<n>To efficiently explore the exponentially growing combination space, AgentOpt implements ten search algorithms.
arXiv Detail & Related papers (2026-04-07T17:13:47Z)
ESearch-R1: Learning Cost-Aware MLLM Agents for Interactive Embodied Search via Reinforcement Learning [40.2017873619555]
ESearch-R1 is a cost-aware embodied reasoning framework.<n>It unifies interactive dialogue (Ask), episodic memory retrieval (GetMemory) and physical navigation (Navigate) into a single decision process.<n>It improves task success rates while reducing total operational costs by approximately 50%.
arXiv Detail & Related papers (2025-12-21T02:45:08Z)
A$^2$FM: An Adaptive Agent Foundation Model for Tool-Aware Hybrid Reasoning [40.6234318894435]
Large language models split into two families: reasoning-centric LLMs and agentic LLMs.<n>This divide arises from fundamentally different training objectives, leading to mismatched strengths and inefficiency on simple queries.<n>We present Adaptive Agent Foundation Model (A$2$FM), a unified framework that follows a route-then-align principle.
arXiv Detail & Related papers (2025-10-13T17:08:25Z)
Learning to Route LLMs from Bandit Feedback: One Policy, Many Trade-offs [69.2486294522259]
BaRP is a Bandit Routing-feedback with Preferences approach that trains under the same partial-feedback restriction as deployment.<n> Framed as a contextual bandit over prompt features and a user preference vector, our method simulates an online feedback setting during training and adapts its routing decisions to each new prompt.
arXiv Detail & Related papers (2025-10-08T18:24:59Z)
Multi-Agent Neural Rewriter for Vehicle Routing with Limited Disclosure of Costs [65.23158435596518]
Solving the multi-vehicle routing problem as a team Markov game with partially observable costs. Our multi-agent reinforcement learning approach, the so-called multi-agent Neural Rewriter, builds on the single-agent Neural Rewriter to solve the problem by iteratively rewriting solutions.
arXiv Detail & Related papers (2022-06-13T09:17:40Z)
Online Apprenticeship Learning [58.45089581278177]
In Apprenticeship Learning (AL), we are given a Markov Decision Process (MDP) without access to the cost function. The goal is to find a policy that matches the expert's performance on some predefined set of cost functions. We show that the OAL problem can be effectively solved by combining two mirror descent based no-regret algorithms.
arXiv Detail & Related papers (2021-02-13T12:57:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.