# LIBRA: Programmatic Induction of Executable Tool Libraries

DevFeed: [LIBRA: Programmatic Induction of Executable Tool Libraries](<https://devfeed.tech/articles/libra-programmatic-induction-of-executable-tool-libraries-118049.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.05445>)

Author: Zekai Li, Zhenke Liu, Yonghao Li, Junhan Liu, Xulong Xiao, Jiapeng Zhang

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Tool](<https://devfeed.tech/topics/tool.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [function calling](<https://devfeed.tech/topics/function-calling.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-retrieval](<https://devfeed.tech/tags/agentic-retrieval.md>), [agents](<https://devfeed.tech/tags/agents.md>), [assembly](<https://devfeed.tech/tags/assembly.md>), [catalog](<https://devfeed.tech/tags/catalog.md>), [contracts](<https://devfeed.tech/tags/contracts.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ir](<https://devfeed.tech/tags/cs-ir.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [framework](<https://devfeed.tech/tags/framework.md>), [induction](<https://devfeed.tech/tags/induction.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [libra](<https://devfeed.tech/tags/libra.md>), [libraries](<https://devfeed.tech/tags/libraries.md>), [library](<https://devfeed.tech/tags/library.md>), [llm](<https://devfeed.tech/tags/llm.md>), [recall](<https://devfeed.tech/tags/recall.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [sandbox](<https://devfeed.tech/tags/sandbox.md>), [similarity](<https://devfeed.tech/tags/similarity.md>), [tool](<https://devfeed.tech/tags/tool.md>)

## AI overview

LIBRA addresses a gap in multi-tool retrieval: ranking tools independently by request similarity can miss workflow prerequisites. Before user queries, an LLM proposes contracts describing candidate invocations, reusable outputs, state changes, and success conditions; LIBRA tests proposed compositions in a sandbox and retains runtime-supported relations. The resulting library needs no dependency annotations, tool-use trajectories, or task-specific training, and supports one-shot and agentic retrieval. Across AppWorld, BFCL, and SciToolEval, it improves one-shot Recall and NDCG over the same dense retriever by average gains of 7.9 and 3.7 percentage points. Within ToolReAGt, gains after ten retrieval iterations are 4.5 and 2.9 points.

## Source excerpt

arXiv:2611.05445v1 Announce Type: new Abstract: Large language model agents increasingly retrieve tools from large catalogs, yet most retrievers rank each tool independently by request similarity. For multi tool tasks, this recovers the goal matching tool while omitting prerequisite tools required by the workflow.We introduce LIBRA, a framework for Library Induction by Runtime Assembly that turns a flat tool catalog into an executable tool library. An LLM proposes per tool contracts of candidate invocations, reusable outputs and state changes, and success conditions; LIBRA then tests these compositions in a sandbox and retains the relations supported at runtime. Constructed before user queries, the library requires no dependency annotations, tool use trajectories, or task specific training, and supports both one-shot and agentic retrieval. On AppWorld, BFCL, and SciToolEval, LIBRA improves one-shot Recall and NDCG over the same dense retriever by 7.9 and 3.7 percentage points on average. Within ToolReAGt, it yields corresponding gains of 4.5 and 2.9 points after ten retrieval iterations.