Daily MT Picks
Archives
Search...
Subscribe
Machine Translation Digest for Oct 21 2026
October 22, 2026
If you are shipping translation systems, the risk is no longer only mistranslation: AI-generated output can be systematically harder to read, with this study...
Machine Translation Digest for Oct 20 2026
October 21, 2026
Tracking an 8.4B-parameter model through three post-training phases on Japanese financial documents, one study finds the stages interact non-additively:...
Machine Translation Digest for Oct 19 2026
October 20, 2026
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...
Machine Translation Digest for Oct 18 2026
October 19, 2026
The biggest practical finding here is that frozen LLM embeddings still encode a language-conditioned geometry even when the sentences mean the same thing, so...
Machine Translation Digest for Oct 15 2026
October 16, 2026
Commercially permissive multilingual SFT and mechanistic work on low-resource translation pull in different directions: one expands what can be trained on,...
Machine Translation Digest for Oct 14 2026
October 15, 2026
Shipping a translation system means more than matching source and target sentences: you have to know whether it keeps media bias intact, follows style and...
Machine Translation Digest for Oct 13 2026
October 14, 2026
AUROC gains that looked solid on two corpora and nine AI-text detectors disappear under a two-sided protocol, where attacks are applied to both human and...
Machine Translation Digest for Oct 12 2026
October 14, 2026
What does it take for an MT system to stay semantically faithful while still landing in the right language, domain, and register? HexaBee-4B tackles that...
Machine Translation Digest for Oct 11 2026
October 12, 2026
For Romanian, Neurograf v1 makes the case for training a language model stack from scratch: 131 billion tokens, a 635 GB corpus, an 80k byte-level BPE...
Machine Translation Digest for Oct 08 2026
October 9, 2026
One paper argues that LLMs can stand in for human judgment on MT quality annotation, while another shows those same models still struggle with hallucination...
Machine Translation Digest for Oct 07 2026
October 8, 2026
Shipping translation systems without knowing where they break is how you end up with fluent output that still misses the job: discourse coherence slips in...
Machine Translation Digest for Oct 06 2026
October 7, 2026
A Jetson Nano–based speech-to-speech pipeline cut weak English→Spanish outputs back into line without retraining, using a Whisper-tiny ASR front end, Opus-...
Machine Translation Digest for Oct 05 2026
October 6, 2026
What exactly are we measuring when we ask a model to “understand” a task: the next token, the right file, the claim’s true status, or the quality of a...
Machine Translation Digest for Oct 04 2026
October 5, 2026
One paper argues that text-to-MQL evaluation should be built by migrating a text-to-SQL benchmark with access-pattern-aware schema design and data-preserving...
Machine Translation Digest for Oct 01 2026
October 2, 2026
One paper argues that a rule-based system still beats fine-tuned neural models for Polish-Silesian, while another takes the opposite tack and tries to...
Machine Translation Digest for Sep 30 2026
October 1, 2026
Shipping translation systems means deciding where consistency, latency, and cost can be bent before they break under real content: long-form subtitles need...
Machine Translation Digest for Sep 29 2026
September 30, 2026
A Greek papyrus HTR pipeline can still support four papyrological tasks even when character error rate is not near zero; in tests on 63,846 current edited...
Machine Translation Digest for Sep 28 2026
September 29, 2026
How do you score a natural-language-to-logic translation when a single quantifier flip can break every proof downstream? One paper answers with SIV, a...
Machine Translation Digest for Sep 27 2026
September 28, 2026
A simple change in loss weighting can make multimodal MT models pay attention to the image when the source text is ambiguous, rather than quietly ignoring...
Machine Translation Digest for Sep 24 2026
September 25, 2026
COILD pushes MT for Indian languages in a more grounded direction, with over 1.16 million human-translated and human-verified sentence pairs across 20...
Older archives
