Yashar Mehdad

Full breakdown + notes: @ymahdad/i-went-to-emnlp-2025-heres-my-reflections-572d07ae15f5" target="_blank" rel="nofollow noopener">medium.com/@ymahdad/i-wen… If you’re building in this space, reply or DM. Happy to trade architectures and practical notes. We need to share more!

Morel of the story: The field needs more plumbers and fewer painters. Fix the leaks, then paint the house.

I went to #EMNLP2025 . Here’s what actually mattered. Short version: smaller/faster models, better retrieval, less agent hype, and a field that needs a cleanup.

We are hiring interns! Are you a PhD student working in multimodal, NLP, computer vision or computer graphics? Are you passionate about generative AI and responsible AI? Please check out these job postings and apply if you're interested!

Interesting work and awesome result by our great intern @alexlimh23

Excited to share our Findings of #EMNLP2022 paper (arxiv.org/abs/2210.13678): Bridging the Training-Inference Gap for Dense Phrase Retrieval. Joint work with amazing collaborators @leejnhk, Barlas Oguz, @XiongWenhan, @icaruszyz, @YasharMehdad, and @WilliamWangNLP. [1/5]

@moustaki @NetflixResearch Congrats!

Quick professional update: after 8 amazing years at @NetflixResearch (time flies!), I started last week at Google, where I'll be focusing on ML for user modeling & personalization -- excited about taking on a new challenge!

Interested in postdoc position at Meta AI (AI Innovation Center) to work on Generative AI? If you have a background in multimodal and CV generation, NeRF models, diffusion networks, and related solutions, with track record of publications in top-tier conferences, contact me here!

[Summarization] Investigating Crowdsourcing Protocols for Evaluating the Factual Consistency of Summaries: lnkd.in/g4WPR_YR

I'm very excited and proud to announce that my team at Meta AI, with our collaborators, will have a strong presence in NAACL 2022 with 8 accepted papers on summarization, question answering and retrieval technologies.#nlproc #ai #NAACL2022 (see papers in follow-up tweets)

[Summarization] CONFIT: Toward Faithful Dialogue Summarization with Linguistically-Informed Contrastive Fine-tuning: lnkd.in/gg3kmRRn

[Summarization and Community QA] AnswerSumm: A Manually-Curated Dataset and Pipeline for Answer Summarization: lnkd.in/g87WR7Qb