Tree-of-Text: A Tree-based Prompting Framework for Table-to-Text Generation in the Sports Domain

arXiv cs.AI / 4/30/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

Tree-of-Text is a tree-structured prompting framework designed to generate accurate sports game reports from structured tables, addressing both interpretation and narrative fluency challenges.
It guides LLMs through three stages: content planning (selecting relevant operations/arguments), operation execution (splitting large tables into sub-tables), and content generation (merging and rewriting short outputs into a cohesive report).
The approach targets common LLM weaknesses in table understanding that can lead to hallucinations under prompt-based table-to-text generation.
Experiments on multiple sports table-to-text datasets show improved performance over prior methods, including better RG/CO metrics on RotoWire-FG and stronger CS/CO results on MLB while using about 40% of the time and cost of Chain-of-Table.
Overall, the paper demonstrates that structured, multi-stage prompting can make table-to-text generation in the sports domain both more effective and more efficient.

Abstract

Generating sports game reports from structured tables is a complex table-to-text task that demands both precise data interpretation and fluent narrative generation. Traditional model-based approaches require large, annotated datasets, while prompt-based methods using large language models (LLMs) often struggle with hallucination due to weak table comprehension. To overcome these challenges, we propose Tree-of-Text, a tree-structured prompting framework that guides LLMs through a three-stage generation process: (1) Content Planning, where relevant operations and arguments are selected from the input tables; (2) Operation Execution, which breaks down large tables into manageable sub-tables; and (3) Content Generation, where short textual outputs are merged and rewritten into a cohesive report. Experiments show that our method outperforms existing methods on ShuttleSet+, leads in RG and CO metrics on RotoWire-FG, and excels in CS and CO on MLB with roughly 40% of the time and cost of Chain-of-Table. These results demonstrate the effectiveness and efficiency of Tree-of-Text and suggest a promising direction for prompt-based table-to-text generation in the sports domain.