# GLEAN: Grounded Lightweight Evaluation Anchors for Contamination-Aware Tabular Reasoning

arXiv: https://arxiv.org/abs/2603.02212 · HTML: https://mlsanity.com/paper/2603.02212/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 36 references that are in the map.
- First posted on arXiv 22 Jan 2026.

## Builds on (most cited first, 23 of 23)

- [DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing](https://mlsanity.com/paper/2111.09543/) (2021)
- [TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second](https://mlsanity.com/paper/2207.01848/) (2022)
- [Compositional Semantic Parsing on Semi-Structured Tables](https://mlsanity.com/paper/1508.00305/) (2015)
- [TabFact: A Large-scale Dataset for Table-based Fact Verification](https://mlsanity.com/paper/1909.02164/) (2019)
- [Snorkel: Rapid Training Data Creation with Weak Supervision](https://mlsanity.com/paper/1711.10160/) (2017)
- [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://mlsanity.com/paper/2107.07653/) (2021)
- [TableLlama: Towards Open Large Generalist Models for Tables](https://mlsanity.com/paper/2311.09206/) (2023)
- [TableBench: A Comprehensive and Complex Benchmark for Table Question Answering](https://mlsanity.com/paper/2408.09174/) (2024)
- [Benchmarking Benchmark Leakage in Large Language Models](https://mlsanity.com/paper/2404.18824/) (2024)
- [WRENCH: A Comprehensive Benchmark for Weak Supervision](https://mlsanity.com/paper/2109.11377/) (2021)
- [TableLLM: Enabling Tabular Data Manipulation by LLMs in Real Office Usage Scenarios](https://mlsanity.com/paper/2403.19318/) (2024)
- [On the Potential of Lexico-logical Alignments for Semantic Parsing to SQL Queries](https://mlsanity.com/paper/2010.11246/) (2020)
- [SCITAB: A Challenging Benchmark for Compositional Reasoning and Claim Verification on Scientific Tables](https://mlsanity.com/paper/2305.13186/) (2023)
- [AntiLeakBench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge](https://mlsanity.com/paper/2412.13670/) (2024)
- [Table-R1: Inference-Time Scaling for Table Reasoning](https://mlsanity.com/paper/2505.23621/) (2025)
- [TableMind: An Autonomous Programmatic Agent for Tool-Augmented Table Reasoning](https://mlsanity.com/paper/2509.06278/) (2025)
- [TableEval: A Real-World Benchmark for Complex, Multilingual, and Multi-Structured Table Question Answering](https://mlsanity.com/paper/2506.03949/) (2025)
- [LastingBench: Defend Benchmarks Against Knowledge Leakage](https://mlsanity.com/paper/2506.21614/) (2025)
- [Tailoring Table Retrieval from a Field-aware Hybrid Matching Perspective](https://mlsanity.com/paper/2503.02251/) (2025)
- [Towards Question Answering over Large Semi-structured Tables](https://mlsanity.com/paper/2502.13422/) (2025)
- [Node-Level Uncertainty Estimation in LLM-Generated SQL](https://mlsanity.com/paper/2511.13984/) (2025)
- [Beyond Query-Level Comparison: Fine-Grained Reinforcement Learning for Text-to-SQL with Automated Interpretable Critiques](https://mlsanity.com/paper/2511.22258/) (2025)
- [What to Keep and What to Drop: Adaptive Table Filtering Framework](https://mlsanity.com/paper/2506.23463/) (2025)

## Abstract (arXiv)

Tabular reasoning benchmarks mix semantic inference, numerical computation, and brittle table formatting, yet evaluations for small models remain vulnerable to contamination, dataset artifacts, and retrieval failures. We propose GLEAN, a lightweight evaluation protocol that integrates contamination-aware probes, weak-supervision governance, retrieval-reasoning diagnostics, and structured error attribution under tight hardware constraints. We evaluate across TabFact, WTQ via Squall, TableBench, RobuT, and SciTab under a 16GB GPU budget. Using Squall gold SQL as an executable anchor (95.2% execution), GLEAN assigns a deterministic error taxonomy (L0-L4 plus L0.5 context miss) and reveals a stable error-mode separation: TAPEX errors skew toward grounding (L3) while TAPAS errors skew toward hallucination/abstention (L2/L0). We validate evidence-row heuristics against SQL-derived rows on simple queries (0.62 precision / 0.71 recall; hybrid recall 0.81) and show that retrieval Recall@K can saturate even when end-to-end EM/F1 remains limited, motivating attribution beyond raw recall. We release a modular framework with audits and sensitivity checks to make small-model tabular evaluation more contamination-aware and diagnostic.