# Anchored Decoding: Provably Reducing Copyright Risk for Any Language Model

Source: [Anchored Decoding: Provably Reducing Copyright Risk for Any Language Model](<https://www.sota2.com/research/paper/698abbb0c97a0b3e3b5de053>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-28T05:00:05.404Z

Authors: Jacqueline He, Jonathan Hayase, Wen-tau Yih, Sewoong Oh, Luke Zettlemoyer, Pang Wei Koh

Published: 2026-02-06T19:00:14.000Z

## Abstract

Language models (LMs) tend to memorize portions of their training data and emit verbatim spans. When the underlying sources are sensitive or copyright-protected, such reproduction raises issues of consent and compensation for creators and compliance risks for developers. We propose Anchored Decoding, a plug-and-play inference-time method for suppressing verbatim copying: it enables decoding from any risky LM trained on mixed-license data by keeping generation in bounded proximity to a permissively trained safe LM. Anchored Decoding adaptively allocates a user-chosen information budget over the generation trajectory and enforces per-step constraints that yield a sequence-level guarantee, enabling a tunable risk-utility trade-off. To make Anchored Decoding practically useful, we introduce a new permissively trained safe model (TinyComma 1.8B), as well as Anchored$\_{\\mathrm{Byte}}$ Decoding, a byte-level variant of our method that enables cross-vocabulary fusion via the ByteSampler framework (Hayase et al., 2025). Across six model pairs on long-form metrics for copying risk and utility, Anchored and Anchored$\_{\\mathrm{Byte}}$ Decoding define a new Pareto frontier, preserving near-original fluency and factuality while closing up to 75% of the measurable copying gap between the risky baseline and a safe reference, at a modest inference overhead.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Factuality on BIOS](<https://www.sota2.com/research/sota/factuality-on-bios.md>): Factuality; BIOS; Factuality: 56.

- [Fluency on BOOKS](<https://www.sota2.com/research/sota/fluency-on-books.md>): Fluency; Books; Fluency Score: 4.46.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2602.07120>)

[PDF](<https://arxiv.org/pdf/2602.07120.pdf>)

[Implementation](<https://github.com/jacqueline-he/anchored-decoding>)