Text generation literature review details tasks, evaluation, and challenges

A systematic literature review published on arXiv analyzes 257 papers on text generation from January 2017 to December 2025. The review categorizes text generation tasks into five main areas: open-ended generation, summarization, translation, paraphrasing, and question answering. It also examines current evaluation methods, including model-free, model-based, and human assessments, and highlights specific challenges such as dataset limitations, coherence issues, and complex reasoning difficulties. The paper further discusses nine common challenges across all tasks, including bias, hallucinations, misuse, privacy, interpretability, transparency, dataset availability, and computational requirements. AI

IMPACT Provides a comprehensive overview of text generation research, aiding researchers in identifying key tasks, evaluation methods, and future research directions.

RANK_REASON The item is a systematic literature review published on arXiv, detailing research on text generation tasks, evaluation, and challenges. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Jonas N. Becker
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →