---
title: "Artificial Epanorthosis: Why Large Language Models Overuse a Classical Rhetorical Figure and How to Mitigate It"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.21498v1"
permalink: "https://feed.astralor.com/items/itm_ee89ef38d82f28630241aed5"
published: "2026-07-23T16:47:39.000Z"
score: 81
channel: "ai"
tags: ["LoRA Adapter","Rhetorical Figure","Large Language Models","RLHF","Text Style Calibration"]
---

# Artificial Epanorthosis: Why Large Language Models Overuse a Classical Rhetorical Figure and How to Mitigate It

## AI commentary

> This paper caught my eye by brilliantly bridging ancient rhetoric and modern LLM training—both scholarly and practical. For AI developers, the one-line instruction and LoRA adapter are ready-to-use tools; for writers and content moderators, it sounds an alarm about the infiltration of machine writing style. Don't just skim the title; the risk warning at the end deserves serious attention from anyone relying on AI-assisted writing.

## At a glance

This paper finds that LLMs systematically overuse the rhetorical figure epanorthosis (self-correction), driven by promotional training data and RLHF preferences, with left-to-right generation as an amplifier. Using an Epanorthosis Index, models overuse in oratory (about twofold, near threefold in Italian) and underuse in informal Q&A. Mitigations include a one-line instruction cutting it by half to three-quarters, and an SFT adapter nearly eliminating it, with a scaling coefficient for calibration to human rates per genre. The goal is calibration, not elimination; the real risk is humans writing like machines.

## Key points

- LLMs systematically overuse epanorthosis, driven by promotional training data and RLHF preferences.
- The Epanorthosis Index reveals overuse in oratory (2x, 3x in Italian) and underuse in informal Q&A.
- Mitigations: a one-line instruction reduces it by half to three-quarters; an SFT adapter nearly eliminates it with scaling to human rate.
- Goal is genre-level calibration to human rate, not elimination; the real risk is humans mimicking machine writing.

## Full text

A rhetorical figure that Cicero and Quintilian catalogued two thousand years ago reappears, systematically, in the text of large language models: epanorthosis, the self-correction of the specimen «This is not a course. It is a journey of transformation». This essay argues that the overuse is a trained disposition, driven mainly by a training distribution rich in promotional prose and by preference tuning (RLHF) that rewards confident, emphatic phrasing; the left-to-right nature of generation is an amplifier rather than the root cause. Building on evidence that models diverge from human rhetorical style, and on Fontanier's classification of epanorthosis as a figure of thought, it sets out a programme that scores the figure against genre-specific human baselines through an Epanorthosis Index (density relative to the human rate). A first measurement, on three sizes of one instruction-tuned model family, finds mis-calibration by register in both directions: the models overshoot in oratory (about twofold, near threefold in Italian, concentrated in the larger tiers) and undershoot in informal question-and-answer writing, while matching humans in argument, journalism, and encyclopedic prose. Three constructive contributions follow: a survey of mitigation techniques centred on lightweight LoRA adapters; a demonstration, in Italian, that a one-line instruction cuts the figure by half to nearly three-quarters and that a supervised-fine-tuning adapter removes it almost entirely, with a scaling coefficient that dials the reduction back onto the human rate; and the argument that the target is calibration to the human rate for each genre, not elimination. It closes on the stakes: the real risk is that we begin to write like the machines.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_ee89ef38d82f28630241aed5