# Understanding the Dynamics of Demonstration Conflict in In-Context Learning

arXiv: https://arxiv.org/abs/2603.04464 · HTML: https://mlsanity.com/paper/2603.04464/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Probing deception, bias and beliefs, one of 13 directions in Mechanistic interpretability.
- Builds on 52 of its 59 references that are in the map.
- First posted on arXiv 3 Mar 2026.

## Placed in

- Mechanistic interpretability / [Probing deception, bias and beliefs](https://mlsanity.com/mechinterp/probing-deception-bias-and-beliefs/)

## Builds on (most cited first, 30 of 52)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?](https://mlsanity.com/paper/2202.12837/) (2022)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [Transformers learn in-context by gradient descent](https://mlsanity.com/paper/2212.07677/) (2022)
- [Larger language models do in-context learning differently](https://mlsanity.com/paper/2303.03846/) (2023)
- [Towards Best Practices of Activation Patching in Language Models: Metrics and Methods](https://mlsanity.com/paper/2309.16042/) (2023)
- [Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts](https://mlsanity.com/paper/2305.13300/) (2023)
- [Trained Transformers Learn Linear Models In-Context](https://mlsanity.com/paper/2306.09927/) (2023)
- [An Empirical Study of GPT-3 for Few-Shot Knowledge-Based VQA](https://mlsanity.com/paper/2109.05014/) (2021)
- [Transformers as Statisticians: Provable In-Context Learning with In-Context Algorithm Selection](https://mlsanity.com/paper/2306.04637/) (2023)
- [Knowledge Conflicts for LLMs: A Survey](https://mlsanity.com/paper/2403.08319/) (2024)
- [Do Llamas Work in English? On the Latent Language of Multilingual Transformers](https://mlsanity.com/paper/2402.10588/) (2024)
- [Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning](https://mlsanity.com/paper/2305.14160/) (2023)
- [Deep Think with Confidence](https://mlsanity.com/paper/2508.15260/) (2025)
- [Scalable Best-of-N Selection for Large Language Models via Self-Certainty](https://mlsanity.com/paper/2502.18581/) (2025)
- [Ground-Truth Labels Matter: A Deeper Look into Input-Label Demonstrations](https://mlsanity.com/paper/2205.12685/) (2022)
- [Found in the Middle: Calibrating Positional Attention Bias Improves Long Context Utilization](https://mlsanity.com/paper/2406.16008/) (2024)
- [Attention Sorting Combats Recency Bias In Long Context Language Models](https://mlsanity.com/paper/2310.01427/) (2023)
- [Overthinking the Truth: Understanding how Language Models Process False Demonstrations](https://mlsanity.com/paper/2307.09476/) (2023)
- [In-Context Learning with Iterative Demonstration Selection](https://mlsanity.com/paper/2310.09881/) (2023)
- [Eliminating Position Bias of Language Models: A Mechanistic Approach](https://mlsanity.com/paper/2407.01100/) (2024)
- [Adversarial Demonstration Attacks on Large Language Models](https://mlsanity.com/paper/2305.14950/) (2023)
- [Meta-in-context learning in large language models](https://mlsanity.com/paper/2305.12907/) (2023)

## Built on by (most cited first, 2 of 2)

- [The Text Uncanny Valley: Non-Monotonic Performance Degradation in LLM Information Retrieval](https://mlsanity.com/paper/2605.07186/) (2026)
- [Understanding Generalization and Forgetting in In-Context Continual Learning](https://mlsanity.com/paper/2605.28705/) (2026)

## Abstract (arXiv)

In-context learning enables large language models to perform novel tasks through few-shot demonstrations. However, demonstrations per se can naturally contain noise and conflicting examples, making this capability vulnerable. To understand how models process such conflicts, we study demonstration-dependent tasks requiring models to infer underlying patterns, a process we characterize as rule inference. We find that models suffer substantial performance degradation from a single demonstration with corrupted rule. This systematic misleading behavior motivates our investigation of how models process conflicting evidence internally. Using linear probes and logit lens analysis, we discover that under corruption models encode both correct and incorrect rules in intermediate layers but develop prediction confidence only in late layers, revealing a two-phase computational structure. We then identify attention heads for each phase underlying the reasoning failures: Vulnerability Heads in early-to-middle layers exhibit positional attention bias with high sensitivity to corruption, while Susceptible Heads in late layers significantly reduce support for correct predictions when exposed to the corrupted evidence. Targeted ablation validates our findings, with masking a small number of identified heads improving performance by over 10%.