Benchmark RadarTRL-Bench
TRL-Bench evaluates tabular encoders at the representation level by exporting row-, column-, or table-level embeddings through each encoder's supported wrapper and probing them with shared lightweight heads across three suites: TRL-CTbench (column/table), TRL-Rbench (row), and TRL-DLTE (compositional Data-Lake Table Enrichment) covering 16 tasks and 20 models.
- Released
- 2026-06-08
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Traditional end-to-end pipelines obscure the comparative quality of tabular encoders from different training paradigms. TRL-Bench provides a standardized protocol to isolate representation-level capability, enabling model selection based on task-specific strengths rather than a single aggregate score.
Motivation
Tabular encoders are usually evaluated inside task-specific end-to-end pipelines, so models from different training paradigms are difficult to compare directly even when they operate on similar tabular signals.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.