AI

Tabular Foundation Models vs Tuned Gradient Boosting for Telecom Churn: A Profit- and Calibration-Aware Evaluation

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 11/11/2026, 05:03:24 UTC.

Briefings

A study evaluated Tabular Foundation Models against tuned Gradient Boosting Trees using five telecom churn datasets. TabPFN-3.5 significantly outperformed tuned GBDTs on Cell2Cell and Iranian datasets in expected maximum profit, while TabICL v2 showed mixed results. The research highlights that licensing costs and split-aware testing are critical factors for production deployment.

Key points

• TabPFN-3.5 achieved significantly higher expected maximum profit than tuned GBDTs on Cell2Cell (1.9%) and Iranian (7.6%) datasets.
• TabICL v2 underpredicted Orange churn by 2.1 percentage points on a 7.3% base rate.

Context

This research compares Tabular Foundation Models against tuned Gradient Boosting Trees on five public telecom churn datasets to evaluate profit and calibration performance.

Limitations

Results are limited to five public datasets and do not account for split-aware testing or production licensing costs.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 11/11/2026, 05:14:24 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeial62ocn3etng4qjqecgsgneiynfmcl2x5pxxg3wj34z46p7wupbe