aiqbench analysis
This research introduces RareCipherFT, a novel training technique that enhances in-context term learning for rare languages by forcing models to recover specific vocabulary through character-shuffled ciphers rather than relying on lexical shortcuts. The method successfully improves term recall across multiple low-resource language pairs without degrading overall translation quality, offering a practical solution for adapting large language models to specialized terminology at deployment.
Training with Ciphers to Enhance In-Context Terminology Learning For Translation from Rare Languages
arXiv:2610.24593v1 Announce Type: cross Abstract: Fine-tuning LLMs with in-context examples (IC-Train) lets translation models adapt to new parallel data at deployment, but it is unclear whether they actually learn the translations of rare terms from those examples. We formalize this capability as in-context term learning (ICTL). We show that standard IC-Train develops this capability only weakly, and propose RareCipherFT, which strengthens it without any lexical resources. RareCipherFT pairs each training target with in-context examples that share a rare term with it. It then replaces that term with a character-shuffled cipher in both the context and the reference, so the term can only be recovered by copying from context. The ciphered fraction of training data is annealed to preserve translation quality. Across seven translation directions between English and Tamil, Assamese, Mizo, Nepali and Hausa, and four models from 1B to 7B parameters, RareCipherFT raises term recall over standard IC-Train by 5-15 absolute points without lowering COMET. Probes using synonyms, nonce words and controls for shared surface context show that these gains come from linking source terms to their in-context translations rather than from shortcuts.
Abstract:Fine-tuning LLMs with in-context examples (IC-Train) lets translation models adapt to new parallel data at deployment, but it is unclear whether they actually learn the translations of rare terms from those examples. We formalize this capability as in-context term learning (ICTL). We show that standard IC-Train develops this capability only weakly, and propose RareCipherFT, which strengthens it without any lexical resources. RareCipherFT pairs each training target with in-context examples that share a rare term with it. It then replaces that term with a character-shuffled cipher in both the context and the reference, so the term can only be recovered by copying from context. The ciphered fraction of training data is annealed to preserve translation quality. Across seven translation directions between English and Tamil, Assamese, Mizo, Nepali and Hausa, and four models from 1B to 7B parameters, RareCipherFT raises term recall over standard IC-Train by 5-15 absolute points without lowering COMET. Probes using synonyms, nonce words and controls for shared surface context show that these gains come from linking source terms to their in-context translations rather than from shortcuts.
Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2610.24593 [cs.CL]
(or arXiv:2610.24593v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.24593 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 274 words
Related coverage
TermJudge: A Document-Level Metric Judging, Not Counting, Terminology in Machine Translation Evaluation
arXiv · Sep 29, 2026
On the Use of LLMs for Specialised Terminology: A Good Alternative to Corpora?
arXiv · Jul 29, 2026
Knowledge Graph Integration With Poro2 For Enriching Medical Text Processing
AMD · Sep 15, 2026
PAMT: Process-Aligned Reinforcement Learning for Multi-Domain Machine Translation
arXiv · Aug 6, 2026