Mining parallel sentences from archival grammars and dictionaries is a very different route to low-resource MT than paraphrasing a question into dozens of rewrites, and that contrast matters here. One paper builds a pipeline for bitext extraction from documentation materials so languages with almost no web presence can still contribute training data; the other studies uncertainty quantification in question answering, using quality-diversity search to generate controlled input perturbations and measure how unstable a model’s answer becomes.

Mobilizing the Archive: Bitext Extraction at Scale from Archival Documents for Extremely Low-resource Languages

Mine archival grammars and dictionaries for bitext, because an OCR-plus-VLM pipeline extracts usable parallel data for 20 languages and gives low-resource MT up to +15.78 chrF2++ in-context.

While mining parallel sentences from web pages has proven effective for low-resource machine translation, the majority of the world's languages have little to no web presence. What they do have, though, are resources that are the result of centuries of efforts by linguists, such as grammars, dictionaries, and other documentation materials. We propose extracting translation examples from such linguistic resources, which offer parallel data for languages under-represented online and could potentially scale to thousands more languages. We develop a pipeline to achieve this, leveraging an Optical Character Recognition (OCR) model as well as a vision language model (VLM), to ensure identified content to be aligned with visual layout information. We demonstrate the utility of the extracted bitexts for bootstrapping translation systems across 20 languages, both by training machine translation models and by applying in-context learning for large language models. For in-context learning, we observe improvements of +15.78 and +6.94 chrF2++ over zero-shot translation for into low- and high-resource languages, respectively.

Neighborhood Elites: Quality-Diversity Search for Uncertainty Quantification in Question Answering

Neighborhood-elites uncertainty sampling is a QA method, but its controlled rewrite space matters for MT evaluation because it separates answer-preserving perturbations by edit type and support explicitness.

Large language models (LLMs) are increasingly capable, yet they produce wrong answers indistinguishable from correct ones. In question answering (QA), such an error propagates unnoticed into downstream tasks. Perturbation-based uncertainty quantification (UQ) addresses this by probing a model over a set of input rewrites and reading uncertainty from how much its predictions vary. Yet this set is usually built generically, by free paraphrasing, token-level perturbation, or adversarial search. Such rewrites leave two things uncontrolled that matter in QA: what kind of change is made to the context, and how explicitly the context supports the answer. When neither is controlled, the composition of the set is determined by the generator rather than by design, and the estimate inherits whatever that generator happens to produce. Quality-diversity (QD) search is built to cover a predefined exploration space, so we recast input perturbation as a QD problem and propose neighborhood elites (NE), a neighborhood constructor that spreads answer-preserving rewrites across two axes, the kind of edit and the explicitness of support. We compare NE against three other neighborhood constructors, holding the editor, QA model, uncertainty measure and search objective fixed, and against three scores read from the original context. Across five extractive-QA datasets and five instruction-tuned LLMs, NE improves failure detection against every other constructor in at least 24 of the 25 dataset-model pairs, reaching 0.726 AUROC and 0.211 AUPRC lift over the base rate. Additionally, NE detects failure better than semantic entropy computed over decoder samples, by 0.074 AUROC and 0.062 AUPRC lift, and this advantage widens as the neighborhood grows.

Papers announced by arXiv on Thursday, 29 October 2026, covering submissions from 28 October to 29 October 2026.