Research map / Large language models / Language, safety and interpretability
LLMs Can Hide Text in Other Text of the Same Length
Antonio Norelli, Michael M. Bronstein
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| hide text e in a stegotext using LLM ranks (Calgacus) | Llama 3 8B · 1000 Reddit posts, 85 tokens | log-probability plausibility of stegotext | Improved |
| compare stegotext with the original hidden text under the LLM | Calgacus stegotexts · 1000 Reddit posts, 85 tokens | text probability | Hurt |
| score stegotexts with a different LLM (Phi-3 3.8B) than the generator (ablation) | Calgacus stegotexts · 1000 Reddit posts, 85 tokens | probability gap between real and fake texts | No clear effect |
| use larger LLM (Gemma 3 27B, Phi4 14B) instead of Llama 3 8B (ablation) | Calgacus · example message and prompt | stegotext soundness | No clear effect |
| use GPT-2 instead of Llama 3 8B (ablation) | Calgacus · standard text | stegotext soundness | Hurt |
| hide a hash (hard-to-predict text) instead of ordinary text (ablation) | Calgacus · example | stegotext coherence | Hurt |
| invert the rank sequence when using a short prompt (ablation) | Calgacus · example | stegotext steering stability | Improved |
| use longer, more detailed secret prompt k (ablation) | Calgacus · example | stegotext steering | Improved |
Closest papers by meaning
- Position: TrustLLM: Trustworthiness in Large Language Models (ICML 2024 · similarity 0.86)
- Beyond Memorization: Violating Privacy via Inference with Large Language Models (ICLR 2024 · similarity 0.86)
- Humanizing the Machine: Proxy Attacks to Mislead LLM Detectors (ICLR 2025 · similarity 0.85)
- Inadvertent Context Leakage in Language Models (NeurIPS 2026 · similarity 0.85)
- Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory (ICLR 2024 · similarity 0.85)
- When Should LLMs Be Less Specific? Selective Abstraction for Reliable Long-Form Text Generation (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
