MateCat worked at a new frontier of Computer Assisted Translation (CAT) technology, that is, how to effectively and ergonomically integrate Machine Translation (MT) within the human translation workflow. At a time in which MT was mainly trained with the objective of creating the most comprehensible output, MateCat targeted the development of MT technology aimed at minimizing the translatorβs post-edit effort. To this end, MateCat developed an enhanced web-based CAT tool offering new MT capabilities, such as automatic adaption to the translated content, online learning from user corrections, and automatic quality estimation.
π€ What Matters in Data for DPO? I asked myself this question a few days ago while trying to understand how to generate a dataset with preferences to run #DPO. This recent #NeurIPS paper answered some of my questions. The findings are simple but crucial for data creation:
π Come and join our group! π
We offer 2 fully funded PhD positions:
π Human-Centred Evaluation Frameworks for Multilingual Technologies (A6)
π€ Multimedia Personalization with Multimodal Large Language Models (A7)
β° Deadline: 15 May 2026
π Details: https://iecs.unitn.it/education/admission/call-for-application
Our pick of the week by
@FBKZhihangXie
: "Detecting Hallucination in SpeechLLMs at Inference Time Using Attention Maps" by @JWaldendorf, Bashar Awwad Shiekh Hasan and Evgenii Tsymbalov
π°
#SpeechLLM #Hallucination
π New paper: Detecting Hallucinations in SpeechLLMs at Inference Time Using Attention Maps
π http://arxiv.org/abs/2604.19565
π§© Lightweight inference-time detection for SpeechLLM hallucinations via audio attention.
β¨ Attention classifiers beat uncertainty baselines on ASR and S2TT.
π New Shared Task: Model Compression for Machine Translation at #WMT2026 (co-located with #EMNLP2026)!
π
Test data out on June 18th, submissions by July 2nd!
Can you shrink an LLM and keep translation quality high? π§ π§
π https://www2.statmt.org/wmt26/model-compression.html #NLP #ML #LLM #ModelCompression