🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

A nice summary for out GPT for MT paper: slator.com/microsoft-larg…

Introducing "KOSMOS-1" as part of our new paper titled "Language Is Not All You Need: Aligning Perception with Language Models" and a step towards "Multimodal LLMs". Paper: arxiv.org/pdf/2302.14045…

Based on @KocmiTom's finding that ChatGPT outperforms existing MT metrics at system-level, we take further steps to investigate few-shot prompts and propose EA Prompting, to significantly improve the segment-level performance. e.g., +10.72 on wmt20 zh-en. github.com/Coldmist-Lu/Er…

Interested in using large language models for translating literary works (e.g., novels)? We hire human translators to annotate errors in LLM translations. In sum, LLMs w/ doc-level context >>> Google Translate, but they still make major errors (across 18 diff language pairs)!

🚨 #SustaiNLP, #ACL2023, #VolunteerReviewersNeeded We need YOUR help to review an overwhelming number of submissions for the #SustaiNLP workshop at #ACL2023. Join our team of reviewers and help us ensure the success of this important workshop the 🌲 ! 🌍! DM me to sign up now.

How do language models deal with ambiguous task specifications? Patterns (biases) start to emerge once we aggregate over a bunch of tasks that are ambiguous by design.

SIGGRAPH 2023 Technical Papers Trailer youtube.com/watch?v=VBZ2sD… via YouTube

At TRAILS, we're hiring a Managing Director, to coordinate research & day-to-day program administration. If you enjoy working with diverse audiences, are organized, and want to work on developing the future of Trustworthy AI, please apply by Sep 11: ejobs.umd.edu/postings/111235

Is your LLaMA still struggling with translation? 🚀ALMA, built on LLaMA-2 with our new MT training recipe, achieves over 12 BLEU & COMET than LLaMA-2's zero-shot and beats all prior works, even NLLB-54B & GPT-3.5 with 7B/13B parameters! Paper: arxiv.org/pdf/2309.11674…

Work done with my amazing teammates: Young Amr Sharaf Hany Awadalla JHU CLSP Code: github.com/fe1ixxu/ALMA Models: huggingface.co/haoranxu

Haoran Xu did an amazing job over the summer as a Microsoft Research intern!

Excited to release ALMA, a fine-tuned LLM model for machine translation, outperforming LLAMA-2 and comparable models on 12 translation directions, joint work lead by Microsoft intern Haoran Xu w/ Young Hany Awadalla

Mistral 7B is out. It outperforms Llama 2 13B on every benchmark we tried. It is also superior to LLaMA 1 34B in code, math, and reasoning, and is released under the Apache 2.0 licence. mistral.ai/news/announcin…

#EMNLP2023 👋Check out our work on Leveraging GPT-4 for Automatic Translation Post-Editing. We formalize the task of automatically post-editing of translations in a generative setting and investigate the efficacy of GPT-4 on the task across multiple axes and language pairs.

Very proud that ALMA has been accepted at ICLR 2024! JHU CLSP Microsoft Research arxiv.org/abs/2309.11674

Opening up a new generation of machine translation leveraging the power of LLMs! It's now in #ICLR2024 (w/ Haoran Xu , Amr Sharaf , Hany Awadalla ). Teaser: Another breakthrough is coming, soonish..

Code: github.com/fe1ixxu/ALMA Model: huggingface.co/haoranxu/ALMA-… Paper: arxiv.org/pdf/2401.08417… Work done with my amazing teammates: Amr Sharaf Yunmo Chen Weiting (Steven) Tan Lingfeng Shen Benjamin Van Durme Kenton Murray Young

📢 I am looking for a student researcher to work with me and my colleagues at Google DeepMind London on understanding & building new neural network architectures. Please reach out to me ([email protected]) and apply below before Mar 22 if interested! google.com/about/careers/…

Excited to share that CPO has been accepted at ICML 2024!