Toward Scalable Verifiable Reward: Proxy State-Based Evaluation for Multi-turn Tool-Calling LLM Agents

• URL: http://arxiv.org/abs/2602.16246v2
• Date: Sun, 22 Feb 2026 05:28:43 GMT
• Title: Toward Scalable Verifiable Reward: Proxy State-Based Evaluation for Multi-turn Tool-Calling LLM Agents
• Authors: Yun-Shiuan Chuang, Chaitanya Kulkarni, Alec Chiu, Avinash Thangali, Zijie Pan, Shivani Shekhar, Yirou Ge, Yixi Li, Uma Kona, Linsey Pang, Prakhar Mehrotra,
• Abstract summary: Large language model (LLM) agents operating via multi-turn dialogue and multi-step tool calling are increasingly used in production.<n>Prior agentic benchmarks rely on fully deterministic backends, which are costly to build and iterate.<n>We propose Proxy State-Based Evaluation, an LLM-driven simulation framework that preserves final state-based evaluation without a deterministic database.
• Score: 8.760287445955045
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Interactive large language model (LLM) agents operating via multi-turn dialogue and multi-step tool calling are increasingly used in production. Benchmarks for these agents must both reliably compare models and yield on-policy training data. Prior agentic benchmarks (e.g., tau-bench, tau2-bench, AppWorld) rely on fully deterministic backends, which are costly to build and iterate. We propose Proxy State-Based Evaluation, an LLM-driven simulation framework that preserves final state-based evaluation without a deterministic database. Specifically, a scenario specifies the user goal, user/system facts, expected final state, and expected agent behavior, and an LLM state tracker infers a structured proxy state from the full interaction trace. LLM judges then verify goal completion and detect tool/user hallucinations against scenario constraints. Empirically, our benchmark produces stable, model-differentiating rankings across families and inference-time reasoning efforts, and its on-/off-policy rollouts provide supervision that transfers to unseen scenarios. Careful scenario specification yields near-zero simulator hallucination rates as supported by ablation studies. The framework also supports sensitivity analyses over user personas. Human-LLM judge agreement exceeds 90%, indicating reliable automated evaluation. Overall, proxy state-based evaluation offers a practical, scalable alternative to deterministic agentic benchmarks for industrial LLM agents.

Related papers

• Claim Automation using Large Language Model [0.0]
  Large Language Models (LLMs) have achieved strong performance on general-purpose language tasks, but their deployment in regulated and data-sensitive domains, including insurance, remains limited.<n>We propose a locally deployed governance-aware language modeling component that generates structured corrective-action recommendations from unstructured claim narratives.<n>We fine-tune pretrained LLMs using Low-Rank Adaptation (LoRA), scoping the model to an initial decision module within the claim processing pipeline to speed up claim adjusters' decisions.
  arXiv  Detail & Related papers  (2026-02-18T20:01:12Z)
• MirrorBench: An Extensible Framework to Evaluate User-Proxy Agents for Human-Likeness [0.4893345190925178]
  Large language models (LLMs) are increasingly used as human simulators.<n> naive "act-as-a-user" often yields verbose, unrealistic utterances.<n>We present MIRRORBENCH, a benchmarking framework that evaluates user proxies solely on their ability to produce human-like user utterances.
  arXiv  Detail & Related papers  (2026-01-13T01:16:13Z)
• How can we assess human-agent interactions? Case studies in software agent design [52.953425368394306]
  We make two major steps towards the rigorous assessment of human-agent interactions.<n>We propose PULSE, a framework for more efficient human-centric evaluation of agent designs.<n>We deploy the framework on a large-scale web platform built around the open-source software agent OpenHands.
  arXiv  Detail & Related papers  (2025-10-10T19:04:28Z)
• CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward [50.97588334916863]
  We develop CompassVerifier, an accurate and robust lightweight verifier model for evaluation and outcome reward.<n>It demonstrates multi-domain competency spanning math, knowledge, and diverse reasoning tasks, with the capability to process various answer types.<n>We introduce VerifierBench benchmark comprising model outputs collected from multiple data sources, augmented through manual analysis of metaerror patterns to enhance CompassVerifier.
  arXiv  Detail & Related papers  (2025-08-05T17:55:24Z)
• MCPEval: Automatic MCP-based Deep Evaluation for AI Agent Models [76.72220653705679]
  We introduce MCPEval, an open-source framework that automates end-to-end task generation and deep evaluation of intelligent agents.<n> MCPEval standardizes metrics, seamlessly integrates with native agent tools, and eliminates manual effort in building evaluation pipelines.<n> Empirical results across five real-world domains show its effectiveness in revealing nuanced, domain-specific performance.
  arXiv  Detail & Related papers  (2025-07-17T05:46:27Z)
• Let's Think in Two Steps: Mitigating Agreement Bias in MLLMs with Self-Grounded Verification [17.67273082468732]
  Verifiers -- functions assigning rewards to agent behavior -- have been key for AI progress in domains like math and board games.<n>We evaluate Multimodal Large Language Models (MLLMs) as verifiers of agent trajectories across web navigation, computer use, and robotic manipulation.<n>We propose Self-Grounded Verification (SGV), a lightweight method that enables more effective use of MLLMs' knowledge and reasoning.
  arXiv  Detail & Related papers  (2025-07-15T18:50:29Z)
• ThinkGeo: Evaluating Tool-Augmented Agents for Remote Sensing Tasks [64.86209459039313]
  ThinkGeo is an agentic benchmark designed to evaluate tool-augmented agents on remote sensing tasks via structured tool use and multi-step planning.<n>We implement a ReAct-style interaction loop and evaluate both open and closed-source LLMs on 486 structured agentic tasks with 1,773 expert-verified reasoning steps.<n>Our analysis reveals notable disparities in tool accuracy and planning consistency across models.
  arXiv  Detail & Related papers  (2025-05-29T17:59:38Z)
• Integrating Counterfactual Simulations with Language Models for Explaining Multi-Agent Behaviour [35.19786322586909]
  We propose Agentic eXplanations via Interrogative Simulation (AXIS)<n>AXIS generates human-centred action explanations for multi-agent policies.<n>We evaluate AXIS on autonomous driving across ten scenarios for five LLMs.
  arXiv  Detail & Related papers  (2025-05-23T12:19:18Z)
• IDA-Bench: Evaluating LLMs on Interactive Guided Data Analysis [60.32962597618861]
  IDA-Bench is a novel benchmark evaluating large language models in multi-round interactive scenarios.<n>Agent performance is judged by comparing its final numerical output to the human-derived baseline.<n>Even state-of-the-art coding agents (like Claude-3.7-thinking) succeed on 50% of the tasks, highlighting limitations not evident in single-turn tests.
  arXiv  Detail & Related papers  (2025-05-23T09:37:52Z)
• TALE: A Tool-Augmented Framework for Reference-Free Evaluation of Large Language Models [16.857263524133284]
  Large Language Models (LLMs) are increasingly integrated into real-world, autonomous applications.<n> relying on static, pre-annotated references for evaluation poses significant challenges in cost, scalability, and completeness.<n>We propose Tool-Augmented LLM Evaluation (TALE), a framework to assess LLM outputs without predetermined ground-truth answers.
  arXiv  Detail & Related papers  (2025-04-10T02:08:41Z)
• TestAgent: Automatic Benchmarking and Exploratory Interaction for Evaluating LLMs in Vertical Domains [19.492393243160244]
  Large Language Models (LLMs) are increasingly deployed in highly specialized vertical domains.<n>Existing evaluations for vertical domains typically rely on the labor-intensive construction of static single-turn datasets.<n>We propose TestAgent, a framework for automatic benchmarking and exploratory dynamic evaluation in vertical domains.
  arXiv  Detail & Related papers  (2024-10-15T11:20:42Z)
• ALI-Agent: Assessing LLMs' Alignment with Human Values via Agent-based Evaluation [48.54271457765236]
  Large Language Models (LLMs) can elicit unintended and even harmful content when misaligned with human values. Current evaluation benchmarks predominantly employ expert-designed contextual scenarios to assess how well LLMs align with human values. We propose ALI-Agent, an evaluation framework that leverages the autonomous abilities of LLM-powered agents to conduct in-depth and adaptive alignment assessments.
  arXiv  Detail & Related papers  (2024-05-23T02:57:42Z)
• Bring Your Own Data! Self-Supervised Evaluation for Large Language Models [52.15056231665816]
  We propose a framework for self-supervised evaluation of Large Language Models (LLMs) We demonstrate self-supervised evaluation strategies for measuring closed-book knowledge, toxicity, and long-range context dependence. We find strong correlations between self-supervised and human-supervised evaluations.
  arXiv  Detail & Related papers  (2023-06-23T17:59:09Z)

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

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.