# Shared Lexical Task Representations Explain Behavioral Variability In LLMs

arXiv: https://arxiv.org/abs/2604.22027 · HTML: https://mlsanity.com/paper/2604.22027/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 48 references that are in the map.
- First posted on arXiv 23 Apr 2026.

## Builds on (most cited first, 21 of 21)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [Dissecting Recall of Factual Associations in Auto-Regressive Language Models](https://mlsanity.com/paper/2304.14767/) (2023)
- [CodeGeeX: A Pre-Trained Model for Code Generation with Multilingual Benchmarking on HumanEval-X](https://mlsanity.com/paper/2303.17568/) (2023)
- [Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers](https://mlsanity.com/paper/2512.15674/) (2025)
- [Summing Up the Facts: Additive Mechanisms Behind Factual Recall in LLMs](https://mlsanity.com/paper/2402.07321/) (2024)
- [The broader spectrum of in-context learning](https://mlsanity.com/paper/2412.03782/) (2024)
- [Training Language Models to Explain Their Own Computations](https://mlsanity.com/paper/2511.08579/) (2025)
- [Belief Dynamics Reveal the Dual Nature of In-Context Learning and Activation Steering](https://mlsanity.com/paper/2511.00617/) (2025)
- [The Dual-Route Model of Induction](https://mlsanity.com/paper/2504.03022/) (2025)
- [The Geometry of Prompting: Unveiling Distinct Mechanisms of Task Adaptation in Language Models](https://mlsanity.com/paper/2502.08009/) (2025)
- [Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition](https://mlsanity.com/paper/2410.05603/) (2024)
- [Do different prompting methods yield a common task representation in language models?](https://mlsanity.com/paper/2505.12075/) (2025)
- [The Heuristic Core: Understanding Subnetwork Generalization in Pretrained Language Models](https://mlsanity.com/paper/2403.03942/) (2024)
- [Do Activation Verbalization Methods Convey Privileged Information?](https://mlsanity.com/paper/2509.13316/) (2025)
- [How Do Language Models Compose Functions?](https://mlsanity.com/paper/2510.01685/) (2025)
- [More or Less Wrong: A Benchmark for Directional Bias in LLM Comparative Reasoning](https://mlsanity.com/paper/2506.03923/) (2025)
- [Mechanism of Task-oriented Information Removal in In-context Learning](https://mlsanity.com/paper/2509.21012/) (2025)

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

- [How Much Do Circuits Tell Us? Measuring the Consistency and Specificity of Language Model Circuits](https://mlsanity.com/paper/2605.08348/) (2026)
- [LLMs Learn Better In-Context from Rules than from Examples](https://mlsanity.com/paper/2609.03213/) (2026)

## Abstract (arXiv)

One of the most common complaints about large language models (LLMs) is their prompt sensitivity -- that is, the fact that their ability to perform a task or provide a correct answer to a question can depend unpredictably on the way the question is posed. We investigate this variation by comparing two very different but commonly-used styles of prompting: instruction-based prompts, which describe the task in natural language, and example-based prompts, which provide in-context few-shot demonstration pairs to illustrate the task. We find that, despite large variation in performance as a function of the prompt, the model engages some common underlying mechanisms across different prompts of a task. Specifically, we identify task-specific attention heads whose outputs literally describe the task -- which we dub lexical task heads -- and show that these heads are shared across prompting styles and trigger subsequent answer production. We further find that behavioral variation between prompts can be explained by the degree to which these heads are activated, and that failures are at least sometimes due to competing task representations that dilute the signal of the target task. Our results together present an increasingly clear picture of how LLMs' internal representations can explain behavior that otherwise seems idiosyncratic to users and developers.