What does it take for an MT system to become controllable without another expensive round of fine-tuning, and how far can LLMs go once the target task is no longer plain translation? GLOSS-MT tackles English–Twi terminology control by substituting source noun phrases with dictionary-backed Twi forms before a single fine-tuning pass on NLLB-200 and related models, while the Moroccan Darija Figurative Language Corpus adds idioms, metaphors and proverbs to evaluate 19 open-weight LLMs on figurative-language translation and MCQ selection. The issue also includes work on a generation-heavy hybrid MT approach, plus two non-MT papers: one that checks whether human-written reasoning steps are actually entailed by their premises, and one that measures how well five LLMs verbalize mathematical formulas across six scientific domains.

GLOSS-MT: Controllable Low-Resource Machine Translation with an Editable Dictionary

For low-resource terminology control, replacing source noun phrases with dictionary translations and fine-tuning once lifts English-to-Asante Twi chrF2 by 2.85-5.48 and writes 79-97% of requested terms.

Adapting a pretrained machine translation (MT) model to new terminology usually requires new parallel data and another round of fine-tuning, which is costly for low-resource languages. We ask whether controllability can be added to MT, and the fine-tuning burden reduced, without degrading the pretrained model. GLOSS-MT looks up the source's noun phrases in an offline English-Twi dictionary, replaces every phrase found by its Twi translation in place, and fine-tunes a translation model once to translate these mixed inputs. We apply the same data and pipeline to NLLB-200 (600M, 1.3B, 3.3B), OPUS-MT en-tw and TranslateGemma 4B. On a document-level English-to-Asante Twi benchmark (84 documents, four domains), every fine-tune scores above its pretrained base in chrF2 (+2.85 to +5.48), and the three NLLB fine-tunes are the best open systems (50.12-50.80). A direct controllability test on human-translated Bible passages shows that the three fine-tunes tested write 79-97% of the dictionary terms, and that replacing every dictionary entry by a different term changes the output accordingly (70-97% of the new terms appear) with no retraining. We release the models, data and code.

Towards Figurative Language Understanding in Moroccan Darija: An Evaluation of Large Language Models

For Moroccan Darija figurative translation, the MDFC benchmark separates literal recognition from generation and shows closed-weight models translate idioms, metaphors, and proverbs best, with context helping unevenly across systems.

Figurative language remains a major challenge for Machine Translation (MT) systems, as the intended meaning of such expressions often goes beyond literal translation. Although recent advances in Large Language Models (LLMs) improved the translation of Arabic dialects, benchmark resources for evaluating figurative language translation in Moroccan Darija remain scarce. To this end, we introduce the Moroccan Darija Figurative Language Corpus (MDFC), a benchmark dataset for MT and multiple-choice question (MCQ) selection covering idioms, metaphors, and proverbs. Using MDFC, we evaluate 19 open- and closed-weight language models and MT systems at both the expression and contextual levels under zero-shot and six-shot settings. We also conduct layer-wise similarity-based probing to determine how model representations distinguish figurative from literal meanings. Our findings reveal that closed-weight models achieve the strongest translation performance, while the benefits of contextual information vary across models and tasks. Differences between translation performance and MCQ accuracy highlight the distinction between generating an appropriate figurative translation and recognizing the intended meaning among candidate translations.

Generation-Heavy Hybrid Machine Translation

For source-poor, target-rich MT, GHMT relies on one translation dictionary plus a source parser and target-language resources, so retargeting new source languages needs far less parallel data than standard systems.

The state of the art techniques in Machine Translation (MT) require large amounts of symmetric resources from source and target languages. This is true regardless of whether the approach is Transfer or Interlingua, Symbolic or Statistical. Symmetry within these approaches is necessary to ensure quality, robustness and retargetability. In reality, such symmetry, whether it is in terms of structural transfer lexicons, interlingual dictionaries or parallel corpora, is a major bottleneck in developing any MT system. This dissertation presents an approach to MT that addresses the lack of symmetry by exploiting symbolic and statistical target language resources in source-poor/target-rich language pairs. This approach is called Generation-Heavy Hybrid Machine Translation (GHMT). Expected source language resources include a syntactic parser and a simple one-to-many translation dictionary. No transfer rules or complex interlingual representations are used. Rich target language symbolic resources are used to overgenerate multiple structural variations from a target-glossed syntactic dependency representation of source language sentences. Statistical target-language resource are then used to select amongst the overgenerated translations. The source-target asymmetry of systems developed in this approach makes them more easily retargetable to new source languages.

Verifying Human-Written Reasoning References

For reasoning evaluation, many human-written reference chains fail entailment only because their cited premise sets are incomplete: 52.3% of 386 derivations pass all checks, and 191 of 230 failed steps become entailed with full premises.

Evaluating the reasoning of AI systems often relies on human-written reasoning chains as references, yet the logical support for their individual steps is rarely verified. Although prior neuro-symbolic approaches have formalized and checked reasoning, their reported evaluations do not systematically assess whether step conclusions in human-written reference derivations are entailed by their cited premise sets or investigate the sources of failures in those references. To address these limitations, we present a novel reference-centred assessment method for human-written derivations that combines entailment checking against cited premise sets, full-premise rechecking, and failure attribution to examine the reliability of these derivations and their formal representations. We evaluate it on P-FOLIO, using a large language model (LLM) to translate step conclusions into first-order logic (FOL) and the Z3 SMT solver to check entailment. Of 386 derivations examined, 202 (52.3%) meet all our checking criteria, meaning that every translated step is entailed by its stated support and the required target is established under the available FOL representations. Among 230 steps not entailed by their stated support, 191 become entailed when checked against the complete original premises, indicating that most initial failures reflect incomplete or mismatched support rather than unsupported claims. Manual analysis of the remaining 39 cases identifies 21 human reasoning errors, 11 FOL translation errors, and 7 dataset annotation errors. A separate manual audit of 50 randomly sampled steps that passed the entailment checks found 46 faithful FOL translations (92%). These findings suggest that disagreement with a human-written reasoning reference may originate in the reference or its formal representation rather than in the AI system being evaluated.

How Well Can LLMs Verbalize Mathematical Formulas? A Multi-Domain Evaluation

For text-to-speech and accessibility, LLMs verbalise formulas unevenly across domains, with GPT-5.2 best on simple transcriptions and extended output remaining much harder, especially without surrounding context.

Large Language Models (LLMs) have shown strong capabilities across natural language tasks. But their ability to generate accurate written verbalizations of mathematical formulas has not been systematically evaluated. Such verbalization is important for text-to-speech systems supporting audio-based learning and visually impaired users. We evaluated five state-of-the-art LLMs across six scientific domains and constructed Read-Aloud Math, a balanced multi-domain corpus with LaTeX formulas, manually validated simple and extended transcriptions, multiple variants, textual context, domain labels, and formula complexity levels. Using multi-reference BLEU, GPT-5.2 achieved the highest score for simple transcriptions (75.9% without context, 64.9% with context), although without context it did not significantly outperform Grok 4.1 Fast or Claude Sonnet 4.5. Claude Sonnet 4.5 achieved the highest score for extended transcriptions without context (36.5%), and Grok 4.1 Fast with context (48.8%). For extended transcriptions, Statistics and Mathematics scored highest across the two context settings, while Physics scored lowest. Formula length was significantly negatively associated with BLEU for simple transcriptions without context, but not with context, whereas BLEU increased significantly with formula length for extended transcriptions in both settings. Grouping symbols were most frequently involved in transcription differences, while variable names, capitalization, and Greek letters were associated with lower BLEU scores. Overall, LLM-based formula verbalization is promising but remains sensitive to transcription type, textual context, and formula characteristics.

Papers announced by arXiv on Monday, 19 October 2026, covering submissions from 16 October to 19 October 2026.