Tag
该论文提出 BRIA（Branch–Inspect–Adopt）智能体，将候选工具的探索隔离在独立分支中并依据执行证据选择性采纳，从而在大型工具库（1897 个工具）上实现更优的工具检索与使用，在 MCP-Atlas 上比 MCP-Zero 提升 14.37 个百分点。
本文提出 Tracer，一种面向 LLM 智能体的轨迹感知工具检索前端，用 trie 索引历史工具调用序列并结合经验与上下文先验对下一步调用进行排序。在六个环境、四个模型上，它在 24 组对比中 21 组提升了工具选择准确率，同时每步输入 token 减少约 61%–80%。
Lookahead-R 将工具检索重构为受资源约束的序贯决策问题，用轻量级执行感知代理世界模型驱动成本敏感的 MCTS，在 ToolBench 的 I3 分割上取得 91.40% 的 NDCG@5，超过 ToolGen (90.16%)。
Cartograph is a federated MCP proxy that enhances AI agent tool discovery by using operator-attested capability cards and a three-layer analysis to reduce token usage and improve retrieval accuracy.
This paper introduces the tool menu as an execution prior and proposes the State-Path Tool Menu framework to improve online agent success rates by learning execution routes for tool retrieval in multi-step tasks.
TRACE introduces a two-stage curriculum to preserve parametric tool knowledge in enterprise LLMs while enabling fast single-beam greedy decoding, achieving improved accuracy and recall over baselines.
ToolSense is an open-source diagnostic framework that generates three benchmarks (realistic retrieval, MCQ probing, QA probing) to audit LLMs' parametric tool knowledge, revealing a knowledge-retrieval dissociation where strong retrieval performance can coexist with poor factual understanding.
BioManus is an MCP-native biomedical agent system that uses graph-scaffolded planning over structured biological capabilities instead of flat prompt-based tool retrieval, achieving better context efficiency and execution accuracy on biomedical benchmarks. The system introduces a BioinfoMCP Compiler to standardize heterogeneous bioinformatics tools and organizes them as a typed heterogeneous MCP graph for scalable reasoning.
CoHyDE introduces an iterative co-training procedure for an LLM rewriter and a dense encoder to improve tool retrieval from large API catalogs. It outperforms single-component baselines, especially on vague queries, by training both components together using InfoNCE and DPO.