Reproduce “Beyond End-to-End Recovery: Towards a Better Understanding of LLM Steganography via Rolewise Decomposition” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Om Tailor, Yucheng Pan, Solaleh Mohammadi, Xinyan Liu, Kaiqing Zhang
The authors evaluate steganographic exchanges by separately measuring whether transmitted text supports deterministic reconstruction, directly exposes the private phrase, and allows a decoder to recover it. Across 7,200 responses, expanded extraction procedures identify 706 additional reconstructable texts missed by original decoders. This changes Gemini-3.5-Flash’s rate from 42.8% to 71.0%; exploratory results do not establish intent or complete verifier coverage.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.