# Mechanism of Task-oriented Information Removal in In-context Learning

arXiv: https://arxiv.org/abs/2509.21012 · HTML: https://mlsanity.com/paper/2509.21012/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Sep 2025.

## Built on by (most cited first, 4 of 4)

- [Emergent Analogical Reasoning in Transformers](https://mlsanity.com/paper/2602.01992/) (2026)
- [Shared Lexical Task Representations Explain Behavioral Variability In LLMs](https://mlsanity.com/paper/2604.22027/) (2026)
- [Noise-Aware In-Context Learning for Hallucination Mitigation in ALLMs](https://mlsanity.com/paper/2604.09021/) (2026)
- [Task Context Shapes Component Roles in Language Models](https://mlsanity.com/paper/2610.13895/) (2026)

## Abstract (arXiv)

In-context Learning (ICL) is an emerging few-shot learning paradigm based on modern Language Models (LMs), yet its inner mechanism remains unclear. In this paper, we investigate the mechanism through a novel perspective of information removal. Specifically, we demonstrate that in the zero-shot scenario, LMs encode queries into non-selective representations in hidden states containing information for all possible tasks, leading to arbitrary outputs without focusing on the intended task, resulting in near-zero accuracy. Meanwhile, we find that selectively removing specific information from hidden states by a low-rank filter effectively steers LMs toward the intended task. Building on these findings, by measuring the hidden states on carefully designed metrics, we observe that few-shot ICL effectively simulates such task-oriented information removal processes, selectively removing the redundant information from entangled non-selective representations, and improving the output based on the demonstrations, which constitutes a key mechanism underlying ICL. Moreover, we identify essential attention heads inducing the removal operation, termed Denoising Heads, which enables the ablation experiments blocking the information removal operation from the inference, where the ICL accuracy significantly degrades, especially when the correct label is absent from the few-shot demonstrations, confirming both the critical role of the information removal mechanism and denoising heads.