Robust Distortion-free Watermarks for Language Models
- URL: http://arxiv.org/abs/2307.15593v3
- Date: Thu, 6 Jun 2024 04:53:25 GMT
- Title: Robust Distortion-free Watermarks for Language Models
- Authors: Rohith Kuditipudi, John Thickstun, Tatsunori Hashimoto, Percy Liang,
- Abstract summary: We propose a methodology for planting watermarks in text from an autoregressive language model.
We generate watermarked text by mapping a sequence of random numbers to a sample from the language model.
- Score: 85.55407177530746
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: We propose a methodology for planting watermarks in text from an autoregressive language model that are robust to perturbations without changing the distribution over text up to a certain maximum generation budget. We generate watermarked text by mapping a sequence of random numbers -- which we compute using a randomized watermark key -- to a sample from the language model. To detect watermarked text, any party who knows the key can align the text to the random number sequence. We instantiate our watermark methodology with two sampling schemes: inverse transform sampling and exponential minimum sampling. We apply these watermarks to three language models -- OPT-1.3B, LLaMA-7B and Alpaca-7B -- to experimentally validate their statistical power and robustness to various paraphrasing attacks. Notably, for both the OPT-1.3B and LLaMA-7B models, we find we can reliably detect watermarked text ($p \leq 0.01$) from $35$ tokens even after corrupting between $40$-$50\%$ of the tokens via random edits (i.e., substitutions, insertions or deletions). For the Alpaca-7B model, we conduct a case study on the feasibility of watermarking responses to typical user instructions. Due to the lower entropy of the responses, detection is more difficult: around $25\%$ of the responses -- whose median length is around $100$ tokens -- are detectable with $p \leq 0.01$, and the watermark is also less robust to certain automated paraphrasing attacks we implement.
Related papers
- Signal Watermark on Large Language Models [28.711745671275477]
We propose a watermarking method embedding a specific watermark into the text during its generation by Large Language Models (LLMs)
This technique not only ensures the watermark's invisibility to humans but also maintains the quality and grammatical integrity of model-generated text.
Our method has been empirically validated across multiple LLMs, consistently maintaining high detection accuracy.
arXiv Detail & Related papers (2024-10-09T04:49:03Z) - Multi-Bit Distortion-Free Watermarking for Large Language Models [4.7381853007029475]
We extend an existing zero-bit distortion-free watermarking method by embedding multiple bits of meta-information as part of the watermark.
We also develop a computationally efficient decoder that extracts the embedded information from the watermark with low bit error rate.
arXiv Detail & Related papers (2024-02-26T14:01:34Z) - On the Learnability of Watermarks for Language Models [80.97358663708592]
We ask whether language models can directly learn to generate watermarked text.
We propose watermark distillation, which trains a student model to behave like a teacher model.
We find that models can learn to generate watermarked text with high detectability.
arXiv Detail & Related papers (2023-12-07T17:41:44Z) - Mark My Words: Analyzing and Evaluating Language Model Watermarks [8.025719866615333]
This work focuses on output watermarking techniques, as opposed to image or model watermarks.
We focus on three main metrics: quality, size (i.e., the number of tokens needed to detect a watermark), and tamper resistance.
arXiv Detail & Related papers (2023-12-01T01:22:46Z) - Improving the Generation Quality of Watermarked Large Language Models
via Word Importance Scoring [81.62249424226084]
Token-level watermarking inserts watermarks in the generated texts by altering the token probability distributions.
This watermarking algorithm alters the logits during generation, which can lead to a downgraded text quality.
We propose to improve the quality of texts generated by a watermarked language model by Watermarking with Importance Scoring (WIS)
arXiv Detail & Related papers (2023-11-16T08:36:00Z) - On the Reliability of Watermarks for Large Language Models [95.87476978352659]
We study the robustness of watermarked text after it is re-written by humans, paraphrased by a non-watermarked LLM, or mixed into a longer hand-written document.
We find that watermarks remain detectable even after human and machine paraphrasing.
We also consider a range of new detection schemes that are sensitive to short spans of watermarked text embedded inside a large document.
arXiv Detail & Related papers (2023-06-07T17:58:48Z) - Watermarking Text Generated by Black-Box Language Models [103.52541557216766]
A watermark-based method was proposed for white-box LLMs, allowing them to embed watermarks during text generation.
A detection algorithm aware of the list can identify the watermarked text.
We develop a watermarking framework for black-box language model usage scenarios.
arXiv Detail & Related papers (2023-05-14T07:37:33Z) - A Watermark for Large Language Models [84.95327142027183]
We propose a watermarking framework for proprietary language models.
The watermark can be embedded with negligible impact on text quality.
It can be detected using an efficient open-source algorithm without access to the language model API or parameters.
arXiv Detail & Related papers (2023-01-24T18:52:59Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.