arXiv cs.CL· Ido Levy, Segev Shlomov, Avi Yaeli, Nir Mashkif, Asaf Adi, Yinon Goldshtein, Eilam Shapira, Leshem Choshen·· 13 小时前AI 评分38
TabAgent：将 LLM 智能体组件蒸馏为廉价分类器
TabAgent: Distilling LLM Agent Components into Cheap Classifiers
AI 导读
TabAgent 提出用基于执行轨迹训练的紧凑文本-表格分类器，替代智能体在路由、候选筛选、门控、验证等闭集决策任务中的 LLM 调用。在 AppWorld 长程基准上，它保持任务级成功率的同时消除候选筛选阶段的 LLM 调用，延迟降低约 95%，推理成本降低 85-91%。v2 版本将评估扩展到三类决策家族的七个智能体基准，TabHead 改为 CatBoost 分类器。
正文 · 原文
Abstract:Agentic systems, AI architectures that autonomously execute multi-step workflows to achieve complex goals, are often built using repeated large language model (LLM) calls for closed-set decision tasks such as routing, shortlisting, gating, and verification. While convenient, this design makes deployments slow and expensive due to cumulative latency and token usage. We propose TabAgent, a framework for replacing generative decision components in closed-set selection tasks with a compact textual-tabular classifier trained on execution traces. TabAgent (i) extracts structured schema, state, and dependency features from trajectories (TabSchema), (ii) augments coverage with schema-aligned synthetic supervision (TabSynth), and (iii) scores candidates with a lightweight classifier (TabHead). On the long-horizon AppWorld benchmark, TabAgent maintains task-level success while eliminating shortlist-time LLM calls, reducing latency by approximately 95% and inference cost by 85-91%. Beyond tool shortlisting, TabAgent generalizes to other agentic decision heads, establishing a paradigm for learned discriminative replacements of generative bottlenecks in production agent architectures.
| Comments: | 67 pages, 14 figures, 26 tables. v2 is a substantial revision with a new title and two added authors. The evaluation is extended from AppWorld to seven agent benchmarks across three decision families (set retrieval, classification, policy and reward), and TabHead is now a CatBoost classifier with an ablation over classical and neural heads |
| Subjects: | Computation and Language (cs.CL) |
| ACM classes: | I.2.6; I.2.7 |
| Cite as: | arXiv:2602.16429 [cs.CL] |
| (or arXiv:2602.16429v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2602.16429 arXiv-issued DOI via DataCite |
Submission history
From: Ido Levy [view email]
[v1]
Wed, 18 Feb 2026 13:01:17 UTC (1,219 KB)
[v2]
Wed, 21 Oct 2026 11:05:33 UTC (828 KB)
来源：arXiv cs.CL · arxiv.org