Another Turn, Better Output? A Turn-Wise Analysis of Iterative LLM Prompting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Javaji, Shashidhar Reddy, Gauri, Bhavul, Zhu, Zining
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911152475209728
author Javaji, Shashidhar Reddy
Gauri, Bhavul
Zhu, Zining
author_facet Javaji, Shashidhar Reddy
Gauri, Bhavul
Zhu, Zining
contents Large language models (LLMs) are now used in multi-turn workflows, but we still lack a clear way to measure when iteration helps and when it hurts. We present an evaluation framework for iterative refinement that spans ideation, code, and math. Our protocol runs controlled 12-turn conversations per task, utilizing a variety of prompts ranging from vague ``improve it'' feedback to targeted steering, and logs per-turn outputs. We score outcomes with domain-appropriate checks (unit tests for code; answer-equivalence plus reasoning-soundness for math; originality and feasibility for ideation) and track turn-level behavior with three families of metrics: semantic movement across turns, turn-to-turn change, and output size growth. Across models and tasks, gains are domain-dependent: they arrive early in ideas and code, but in math late turns matter when guided by elaboration. After the first few turns, vague feedback often plateaus or reverses correctness, while targeted prompts reliably shift the intended quality axis (novelty vs. feasibility in ideation; speed vs. readability in code; in math, elaboration outperforms exploration and drives late-turn gains). We also observe consistent domain patterns: ideation moves more in meaning across turns, code tends to grow in size with little semantic change, and math starts fixed but can break that path with late, elaborative iteration. Together, the framework and metrics make iteration measurable and comparable across models, and signal when to steer, stop, or switch strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2509_06770
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Another Turn, Better Output? A Turn-Wise Analysis of Iterative LLM Prompting
Javaji, Shashidhar Reddy
Gauri, Bhavul
Zhu, Zining
Artificial Intelligence
Human-Computer Interaction
Large language models (LLMs) are now used in multi-turn workflows, but we still lack a clear way to measure when iteration helps and when it hurts. We present an evaluation framework for iterative refinement that spans ideation, code, and math. Our protocol runs controlled 12-turn conversations per task, utilizing a variety of prompts ranging from vague ``improve it'' feedback to targeted steering, and logs per-turn outputs. We score outcomes with domain-appropriate checks (unit tests for code; answer-equivalence plus reasoning-soundness for math; originality and feasibility for ideation) and track turn-level behavior with three families of metrics: semantic movement across turns, turn-to-turn change, and output size growth. Across models and tasks, gains are domain-dependent: they arrive early in ideas and code, but in math late turns matter when guided by elaboration. After the first few turns, vague feedback often plateaus or reverses correctness, while targeted prompts reliably shift the intended quality axis (novelty vs. feasibility in ideation; speed vs. readability in code; in math, elaboration outperforms exploration and drives late-turn gains). We also observe consistent domain patterns: ideation moves more in meaning across turns, code tends to grow in size with little semantic change, and math starts fixed but can break that path with late, elaborative iteration. Together, the framework and metrics make iteration measurable and comparable across models, and signal when to steer, stop, or switch strategies.
title Another Turn, Better Output? A Turn-Wise Analysis of Iterative LLM Prompting
topic Artificial Intelligence
Human-Computer Interaction
url https://arxiv.org/abs/2509.06770