# SICI: A Semantic-Pragmatic Complexity Index Reveals Regime Shifts in LLM Stance Detection

arXiv: https://arxiv.org/abs/2606.13189 · HTML: https://mlsanity.com/paper/2606.13189/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 32 references that are in the map.
- First posted on arXiv 11 Jun 2026.

## Builds on (most cited first, 8 of 8)

- [BERTweet: A pre-trained language model for English Tweets](https://mlsanity.com/paper/2005.10200/) (2020)
- [Zero-Shot Stance Detection: A Dataset and Model using Generalized Topic Representations](https://mlsanity.com/paper/2010.03640/) (2020)
- [Stance Detection on Social Media: State of the Art and Trends](https://mlsanity.com/paper/2006.03644/) (2020)
- [Adversarial Learning for Zero-Shot Stance Detection on Social Media](https://mlsanity.com/paper/2105.06603/) (2021)
- [Improved Target-specific Stance Detection on Social Media Platforms by Delving into Conversation Threads](https://mlsanity.com/paper/2211.03061/) (2022)
- [Chain-of-Thought Embeddings for Stance Detection on Social Media](https://mlsanity.com/paper/2310.19750/) (2023)
- [Stance Reasoner: Zero-Shot Stance Detection on Social Media with Explicit Reasoning](https://mlsanity.com/paper/2403.14895/) (2024)
- [Are Stereotypes Leading LLMs' Zero-Shot Stance Detection ?](https://mlsanity.com/paper/2510.20154/) (2025)

## Abstract (arXiv)

Prompt-based LLMs are increasingly used for stance detection, but harder examples are not always repaired by clearer instructions, reasoning prompts, retrieval, or debate. We introduce SICI (Stance Inference Complexity Index), a seven-dimensional diagnostic measure of the semantic-pragmatic burden imposed by a target--text pair. Across SemEval-2016 and VAST, SICI predicts LLM accuracy better than surface proxies and shows substantial cross-scorer reliability (α=0.771). More importantly, LLM errors change regime as SICI increases: low-complexity examples invite over-attribution, especially Against predictions; intermediate examples form an unstable boundary; and high-complexity examples rapidly concentrate on None. This phase-transition-like structure persists across GPT-3.5, GPT-4o-mini, DeepSeek-V3, and GPT-4o, although stronger models move the boundaries. A 15-method intervention study further shows that prompting, retrieval, and debate often shift models along the attribution--abstention axis rather than removing the high-complexity bottleneck.