# FERN: Cross-Benchmark Table Retrieval for Text-to-SQL with Frozen Retrievers and Learned Ranking

Source: [arXiv](https://arxiv.org/abs/2611.14477v1)  
Feed7 permalink: https://feed7.dev/p/2611-14477v1-01roe2s  
Published: 2026-11-10T18:39:49.000Z  
Trust: Needs Review (needs_review)

## Why Included

FERN ranks database tables for text-to-SQL without generative model calls at retrieval time. Its results show strong gains on some benchmarks, but weaker transfer on others makes dataset-specific validation essential.

## Source Summary

FERN combines three frozen retrievers, describes up to **120 features** for each of 100 candidates, and ranks them with LightGBM fitted on BIRD training questions. On Spider, complete@5 rose from **83.9% to 95.6%**; median latency on BIRD dev was **208 ms** per question on one A100, without hosted model calls.

## Practical Implication

Builders retrieving schema context for text-to-SQL can test a learned reranker between retrieval and generation. It may improve which tables reach the model while keeping retrieval independent of hosted LLM calls.

## Agent-Ready Context

FERN combines three frozen retrievers, describes up to **120 features** for each of 100 candidates, and ranks them with LightGBM fitted on BIRD training questions. On Spider, complete@5 rose from **83.9% to 95.6%**; median latency on BIRD dev was **208 ms** per question on one A100, without hosted model calls.

Builders retrieving schema context for text-to-SQL can test a learned reranker between retrieval and generation. It may improve which tables reach the model while keeping retrieval independent of hosted LLM calls.

Transfer was uneven: on Fiben, recall at fifteen tables was 54.2% versus RASL’s 69.2%, using different question sets. FERN’s BEAVER complete@15 also trailed two BEAVER-trained retrievers, so results from one catalog should not be assumed to generalize.

## Connected Context

Feed7 judgment across 1059 accumulated Signals:

FERN shows that table retrieval can benefit from a learned ranking stage over frozen retrievers, but its uneven transfer limits any general claim that this ranking will work across catalogs. The practical implication is to evaluate retrieval and ranking on the target schema and question distribution.

- [Retriever-Free Retrieval: Navigating Documents with Decision Models](https://feed7.dev/p/2610-35868v1-1uzxrmk) — FERN reranks candidates produced by fixed retrievers, while DecisionRAG changes the search process by navigating document branches without a first-stage retriever. They are alternative retrieval designs to compare for a given corpus.
- [Graphify-Labs/graphify](https://feed7.dev/p/graphify-1e0bs1f) — FERN selects schema tables for SQL generation; Graphify retrieves project relationships for coding agents. Both make structured context selection explicit, though their data and ranking needs differ.

## Context Map

- Layer: context
- Domains: coding, data
- Topics: retrieval, context-engineering

## Uncertainty

- Transfer was uneven: on Fiben, recall at fifteen tables was 54.2% versus RASL’s 69.2%, using different question sets. FERN’s BEAVER complete@15 also trailed two BEAVER-trained retrievers, so results from one catalog should not be assumed to generalize.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.
