MAVEN: Improving Generalization in Agentic Tool Calling
Quick Answer
This paper shows that MAVEN (Modular Agentic Verification and Execution Network) enhances reasoning in agentic tool-calling environments, improving GPT-OSS-120b accuracy from 48% to 71% on MAVEN-Bench without extra training.
Quick Take
This lightweight framework also remains competitive against proprietary models at a cost ratio of 1/10, highlighting its potential for better compositional reasoning.
Key Points
- MAVEN introduces a symbolic reasoning scaffold for structured decomposition and adaptive tool orchestration.
- Evaluated across benchmarks like BFCL v3 and AceBench, MAVEN-Bench tests multi-step reasoning.
- MAVEN's improvements demonstrate a significant gap in reasoning quality versus task success.
- The model achieves a 23% accuracy increase without additional training, showcasing efficiency.
- MAVEN's cost-effectiveness suggests a new approach for process-aware .
Paper Resources
Article Content
From source RSS / original summaryarXiv:2605. 30738v1 Announce Type: new Abstract: Generalization across agentic tool-calling environments remains a central challenge for reliable agentic reasoning systems. Although achieve strong results on individual benchmarks, their ability to compose reasoning strategies, preserve intermediate states, and coordinate tools across domains remains underexplored.
We present MAVEN (Modular Agentic Verification and Execution Network), a lightweight symbolic reasoning scaffold for structured decomposition, adaptive tool orchestration, and intermediate verification. We evaluate MAVEN across established tool-calling benchmarks, including BFCL v3, TauBench, Tau2Bench, AceBench, and introduce MAVEN-Bench, a stress-test benchmark for multi-step mathematical and physical reasoning with explicit verification and adversarial task composition.
MAVEN-Bench exposes a substantial gap between partial reasoning quality and end-to-end task success; in direct MAVEN-Bench runs, MAVEN improves its GPT-OSS-120b base model from 48% to 71% accuracy without additional training.
It also remains competitive with frontier proprietary baselines while using an open-weight backbone with an estimated cost ratio of roughly 1/10, suggesting that lightweight verification-centered scaffolds can strengthen compositional reasoning and motivate more process-aware evaluation of agents in the wild.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.