CAST: Achieving Stable LLM-based Text Analysis for Data Analytics
Not on the map. Why?
Abstract (arXiv)
Text analysis of tabular data relies on two core operations: summarization for corpus-level theme extraction and tagging for row-level labeling. A critical limitation of employing large language models (LLMs) for these tasks is their inability to meet the high standards of output stability demanded by data analytics. To address this challenge, we introduce CAST (Consistency via Algorithmic Prompting and Stable Thinking), a framework that enhances output stability by constraining the model's latent reasoning path. CAST combines (i) Algorithmic Prompting to impose a procedural scaffold over valid reasoning transitions and (ii) Thinking-before-Speaking to enforce explicit intermediate commitments before final generation. To measure progress, we introduce CAST-S and CAST-T, stability metrics for bulleted summarization and tagging, and validate their alignment with human judgments. Experiments across publicly available benchmarks on multiple LLM backbones show that CAST consistently achieves the best stability among all baselines, improving Stability Score by up to 16.2%, while maintaining or improving output quality.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 19 of its 63 references that are in the map. First posted on arXiv 26 Jan 2026.
Builds on
19 of 63 references.
References · 19
- 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
- 2019BERTScore: Evaluating Text Generation with BERTcited by 5,512
- 2020SummEval: Re-evaluating Summarization Evaluationcited by 674
- 2024Justice or Prejudice? Quantifying Biases in LLM-as-a-Judgecited by 368
- 2023StructGPT: A General Framework for Large Language Model to Reason over Structured Datacited by 343
- 2024Quiet-STaR: Language Models Can Teach Themselves to Think Before Speakingcited by 301
- 2022MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languagescited by 146
- 2022Revisiting the Gold Standard: Grounding Summarization Evaluation with Robust Human Evaluationcited by 142
- 2023Decomposing Uncertainty for Large Language Models through Input Clarification Ensemblingcited by 114
- 2022Teaching Algorithmic Reasoning via In-context Learningcited by 112
- 2024FineSurE: Fine-grained Summarization Evaluation using LLMscited by 83
- 2024Factual Confidence of LLMs: on Reliability and Robustness of Current Estimatorscited by 61
7 more
- 2023Training Chain-of-Thought via Latent-Variable Inferencecited by 47
- 2020Towards Interpretable Natural Language Understanding with Explanations as Latent Variablescited by 30
- 2023ReConTab: Regularized Contrastive Representation Learning for Tabular Datacited by 25
- 2024StrucText-Eval: Evaluating Large Language Model's Reasoning Ability in Structure-Rich Textcited by 13
- 2023Think Before You Speak: Cultivating Communication Skills of Large Language Models via Inner Monologuecited by 11
- 2025SALMAN: Stability Analysis of Language Models Through the Maps Between Graph-based Manifoldscited by 2
- 2019A framework for streamlined statistical prediction using topic modelscited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2023Rethinking Tabular Data Understanding with Large Language Models
- 2025How well do LLMs reason over tabular data, really?
- 2025RUST-BENCH: Benchmarking LLM Reasoning on Unstructured Text within Structured Tables
- 2025Utilizing Training Data to Improve LLM Reasoning for Tabular Understanding
- 2025RealHiTBench: A Comprehensive Realistic Hierarchical Table Benchmark for Evaluating LLM-Based Table Analysis
- 2024Tree-of-Table: Unleashing the Power of LLMs for Enhanced Large-Scale Table Understanding
- 2025NeedleInATable: Exploring Long-Context Capability of Large Language Models towards Long-Structured Tables
- 2026Why Large Language Models Fail at Tabular Prediction