Unsupervised Tool Evolution for Coding Agents in Data-Lake Analytics
We propose TEAL (Tool Evolution for Analytics over data Lakes), which evolves a library of tools and usage skills for a given Worker agent without labels.
ProofPaper ↗
Key points
- Coding agents can answer analytical questions over a data lake by discovering, cleaning, and joining the relevant data.
- Since a ground-truth label for a data-lake question requires an expert to solve it first, we consider the unsupervised setting.
- Two agents propose tools from complementary evidence, an Explorer from the data lake itself and a Critic from Worker trajectories.
- We also introduce TEAL-Q, which generates synthetic questions for tool evolution when real ones are unavailable.
Sources (2)
- [1]Unsupervised Tool Evolution for Coding Agents in Data-Lake AnalyticsHugging Face Daily Papers · Oct 19, 12:00 AM
We propose TEAL (Tool Evolution for Analytics over data Lakes), which evolves a library of tools and usage skills for a given Worker agent without labels.
Coding agents can answer analytical questions over a data lake by discovering, cleaning, and joining the relevant data.
- [2]Unsupervised Tool Evolution for Coding Agents in Data-Lake AnalyticsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 09:44 PM · same content
Extractive summary: sentences quoted from the sources.
Before this
- Oct 18, 2026Bridging Concepts: Multi-Agent Reasoning over Scientific Concept Graphs for High-Impact Research Prediction
- Oct 18, 2026Toward Agentic Retrieval-Augmented Generation for Enterprise Retail Workflows over Heterogeneous Data Sources
- Oct 18, 2026SoK: PIAtlas: Decomposing the Unbounded Prompt Injection Attack Space
- Oct 17, 2026Refuse the Harm, Complete the Task via Graph-of-Rules Planning for LLM-Enabled Robots
- Oct 16, 2026sparqlite: SPARQL 1.1 inside SQLite, written in Rust
- Oct 16, 2026OR-License: Evaluating LLM Agents as Real-World Operations Research Practitioners