# Semantic Overlays: Mitigating Prompt Injection with Annotations Beyond Tokens and Steering Vectors

arXiv: https://arxiv.org/abs/2608.23873 · HTML: https://mlsanity.com/paper/2608.23873/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Prompt injection and agent security, one of 12 directions in LLM safety and security.
- Builds on 19 of its 21 references that are in the map.
- First posted on arXiv 24 Aug 2026.

## Placed in

- LLM safety and security / [Prompt injection and agent security](https://mlsanity.com/safety/prompt-injection-and-agent-security/)
- Mechanistic interpretability / [Steering and probing for LLM safety](https://mlsanity.com/mechinterp/steering-and-probing-for-llm-safety/)

## Builds on (most cited first, 19 of 19)

- [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://mlsanity.com/paper/2101.00190/) (2021)
- [The Power of Scale for Parameter-Efficient Prompt Tuning](https://mlsanity.com/paper/2104.08691/) (2021)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions](https://mlsanity.com/paper/2404.13208/) (2024)
- [StruQ: Defending Against Prompt Injection with Structured Queries](https://mlsanity.com/paper/2402.06363/) (2024)
- [Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet](https://mlsanity.com/paper/2605.29358/) (2026)
- [Tensor Trust: Interpretable Prompt Injection Attacks from an Online Game](https://mlsanity.com/paper/2311.01011/) (2023)
- [Can LLMs Separate Instructions From Data? And What Do We Even Mean By That?](https://mlsanity.com/paper/2403.06833/) (2024)
- [Instructional Segment Embedding: Improving LLM Safety with Instruction Hierarchy](https://mlsanity.com/paper/2410.09102/) (2024)
- [ASIDE: Architectural Separation of Instructions and Data in Language Models](https://mlsanity.com/paper/2503.10566/) (2025)
- [A Comparative Study of Programming Languages in Rosetta Code](https://mlsanity.com/paper/1409.0252/) (2014)
- [PIArena: A Platform for Prompt Injection Evaluation](https://mlsanity.com/paper/2604.08499/) (2026)
- [Stronger Enforcement of Instruction Hierarchy via Augmented Intermediate Representations](https://mlsanity.com/paper/2505.18907/) (2025)
- [Steering Instruction Hierarchies at Inference Time](https://mlsanity.com/paper/2607.26228/) (2026)
- [Epistemic Goggles: A Pretrained Module that Induces an Epistemic Frame via Gradient Editing](https://mlsanity.com/paper/2607.01690/) (2026)

## Abstract (arXiv)

Everything a language model sees is tokens. The serving stack knows what each span is -- user input, tool output, instructions -- but the model must keep track of that itself, and can lose track or be confused: text can be written to read like anything. Prompt injection is a natural exploit of this phenomenon. By scrambling the model's understanding of span identity, an attacker can induce unwanted and dangerous actions. Adding a non-textual channel to the model's input -- a way to communicate span identity beyond text -- mitigates this class of attack. We thus introduce a general steering technique called Semantic Overlays: small learned adapters applied at chosen prefill positions to a frozen model's residual stream. Laying an overlay over a span creates an out-of-band annotation channel that cannot be replicated by tokens. Unlike steering vectors, Semantic Overlays are trained, adaptable, and selectively applied. An overlay can encode complex semantics that reshape how the model perceives the marked span: asked to copy a code snippet under an overlay asserting a different programming language, the model rewrites the snippet in the asserted language. Overlays compose, allow transparent reading of underlying content, and can carry complex payloads -- including imperatives the model will follow. An overlay which marks a span as "non-executable" defends against the broad class of prompt injections that add instructions in untrusted context. We report strong results on five prompt injection benchmarks: SEP separation rises from 24.3% to 99.0% with utility unchanged (our scoring rule; we correct a defect in the published grader), TensorTrust attack success falls from 34.8% to 6.2%, AlpacaFarm from 99.0% to 0%, and the overlay beats every published PIArena defense that leaves the model able to answer -- while marked spans stay readable, all at >95% character similarity to the original.