The Rise of Verbal Tics in Large Language Models: A Systematic Analysis Across Frontier Models

2026-04-21 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

A systematic analysis investigates the emergence of verbal tics in Large Language Models (LLMs), defined as repetitive, formulaic linguistic patterns like sycophantic openers or overused vocabulary. The study evaluates eight frontier LLMs, including GPT-5.4, Claude Opus 4.7, and Gemini 3.1 Pro, using a custom API-based framework across 10,000 prompts in 10 task categories in both English and Chinese, generating 160,000 responses. Researchers introduce the Verbal Tic Index (VTI) to quantify tic prevalence, finding significant inter-model variation, with Gemini 3.1 Pro exhibiting the highest VTI (0.590) and DeepSeek V3.2 the lowest (0.295). Verbal tics accumulate in multi-turn conversations, are amplified in subjective tasks, and show cross-lingual patterns. Human evaluation (N = 120) confirms a strong inverse correlation between sycophancy and perceived naturalness (r = -0.87, p < 0.001).

Key takeaway

For AI Product Managers designing conversational agents, understanding the "alignment tax" of verbal tics is crucial. Your models' perceived naturalness directly correlates with lower sycophancy and fewer repetitive phrases. Prioritize LLMs with lower Verbal Tic Index scores, like DeepSeek V3.2, and implement strategies to mitigate tic accumulation in multi-turn interactions to enhance user experience and trust.

Key insights

LLMs exhibit pervasive verbal tics, an "alignment tax" impacting perceived naturalness and accumulating in multi-turn interactions.

Principles

• Verbal tics correlate inversely with perceived naturalness.
• Tics increase in multi-turn conversations and subjective tasks.

Method

A custom evaluation framework assesses 10,000 prompts across 10 task categories in English and Chinese, generating 160,000 responses. The Verbal Tic Index (VTI) quantifies tic prevalence.

In practice

• Monitor VTI in LLM outputs.
• Prioritize models with lower VTI for naturalness.
• Address tic accumulation in conversational AI.

Topics

• Verbal Tics
• Large Language Models
• LLM Alignment
• Verbal Tic Index
• Model Evaluation

Best for: AI Engineer, Research Scientist, AI Product Manager, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• The Rise of Verbal Tics in Large Language Models: A Systematic Analysis Across Frontier Models — LLMs exhibit pervasive verbal tics, an "alignment tax" impacting naturalness and trust due to current training paradigms.
• Probing Semantic Alignment, Lexical Invariance, and Syntactic Influence in LLM Metaphor Processing — LLMs struggle with deep metaphor comprehension, often misinterpreting concepts and relying on lexical associations over context or full syntactic understanding.
• Fully Automated Identification of Lexical Alignment and Preference-Stage Shifts in Large Language Models — Automated metrics can identify and quantify lexical overuse and preference learning shifts in LLMs without manual curation.
• ChatGPT Has ‘Goblin’ Mania in the US. In China It Will ‘Catch You Steadily’ — ChatGPT develops distinct, sometimes frustrating, verbal tics in both English and Chinese.
• As Easy as Rocket Science: Assessing the Ability of Large Language Models to Interpret Negation in Figurative Language — LLMs struggle with combined negation and figurative language, with prompt style significantly impacting performance.
• Revisiting the Systematicity in Negation in the Era of In-Context Learning — LLMs partially grasp negation via in-context learning, but struggle with robust scope recognition and function vector composition for it.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.