Selected primary research
Methods, guarantees, reliability, and related uses of generated text.
Ordered by earliest documented date. The paper pages distinguish manuscript dates, first preprints, and published versions and explain each work’s relationship to Neural Linguistic Steganography. This is a curated starting point, not an exhaustive survey.
10 papers · all entries are available without JavaScript
Uses an LSTM language model and vocabulary bins to generate text carrying a secret bitstream. Experiments on tweets and email explore the trade-off between embedding capacity and text quality.
First arXiv submission: May 30, 2017 · ACL 2017 Student Research Workshop
Generates cover text with a recurrent language model and dynamically encodes candidate words using fixed-length or Huffman variable-length codes. It evaluates generation efficiency, embedding capacity, text quality, and resistance to the tested steganalysis methods.
First online publication (IEEE): September 24, 2018 · IEEE TIFS 2019 (online 2018)
Analyzes how binning and variable-length encoding alter a language model's output distribution. Its patient-Huffman method skips message embedding at steps where the induced distribution would differ too much from the model.
First arXiv submission: July 15, 2019 · ACL 2019
Combines arithmetic coding with large pretrained language models to encode secret information into generated text while closely matching the model's distribution. GPT-2 experiments evaluate information capacity, statistical divergence, and human judgments of contextual continuations, and demonstrate encoding natural-language messages into unrelated cover text.
First arXiv submission: September 3, 2019 · EMNLP-IJCNLP 2019
Introduces self-adjusting arithmetic coding, choosing the candidate-set size at each generation step to control divergence from the language model. It supplies an imperceptibility analysis and evaluates capacity, divergence, and human judgments across multiple datasets.
First arXiv submission: October 1, 2020 · EMNLP 2020
Characterizes perfect steganographic security through couplings and maximal information throughput through minimum entropy couplings under its information-theoretic model. It develops an iterative method and compares it with arithmetic coding, Meteor, and adaptive dynamic grouping on text, audio, and image-model distributions.
First arXiv submission: October 24, 2022 · ICLR 2023
Embeds a statistical watermark by softly favoring a pseudorandom set of tokens during generation. It develops a detection test that does not require the language model and analyzes how the token distribution affects detection strength.
First arXiv submission: January 24, 2023 · ICML 2023
Formalizes secret collusion among generative AI agents and develops capability evaluations for covert communication and coordination. It studies individual and multi-agent settings alongside the assumptions and limitations of monitoring and other mitigations.
First arXiv submission: February 12, 2024 · NeurIPS 2024
Introduces SyncPool to resolve subword segmentation ambiguity while preserving the model's token distribution. It groups prefix-related tokens and uses synchronized pseudorandom sampling so a receiver can recover token choices from detokenized text.
First arXiv submission: March 26, 2024 · IEEE TDSC 2025 (preprint 2024)
Studies tokenization inconsistencies between generated token sequences and the text a receiver retokenizes. It proposes stepwise verification for steganography and post-hoc rollback for watermark detection, evaluating how these changes affect reliability, text quality, and detection.
First arXiv submission: August 28, 2025 · EMNLP 2025
No papers match those filters. Try a broader term or reset the filters.