Victoria X Lin (@VictoriaLinML) / X

Victoria X Lin

1,447 posts

Victoria X Lin

@VictoriaLinML

MTS @thinkymachines | Native Multimodal Intelligence Prev: @AIatMeta @SFResearch • PhD @uwcse

San Francisco Bay Area

Joined December 2010

Pinned
Victoria X Lin
@VictoriaLinML
May 11
✨We are showing some experiments with interaction models @thinkymachines: models that could see and hear continuously while processing tasks in the background and generating responses in real-time. Interaction models offer a glimpse into a future where people collaborate with
Thinking Machines
@thinkymachines
May 11
People talk, listen, watch, think, and collaborate at the same time, in real time. We've designed an AI that works with people the same way. We share our approach, early results, and a quick look at our model in action. thinkingmachines.ai/blog/interacti…
00:00
11K
Victoria X Lin
@VictoriaLinML
Oct 4, 2023
🌐 Retrieval-augmentation is becoming increasingly popular for enriching LLMs with long-tail knowledge and keeping them up-to-date. 💫 Introducing RA-DIT: Retrieval-Augmented Dual Instruction Tuning (arxiv.org/abs/2310.01352), a fine-tuning process that enhances LM and retrieval
75K
Victoria X Lin
@VictoriaLinML
Aug 1, 2024
1/n Introducing MoMa 🖼, our new sparse early-fusion architecture for mixed-modal language modeling that significantly boosts pre-training efficiency 🚀 (arxiv.org/pdf/2407.21770). MoMa employs a mixture-of-expert (MoE) framework with modality-specific expert groups. Given any
102K
Victoria X Lin
@VictoriaLinML
Nov 9, 2024
💫 Introducing Mixture-of-Transformers (MoT) , our latest work advancing modality-aware sparse architectures for multimodal foundation models, led by @liang_weixin, in collaboration w/ amazing colleagues at @AIatMeta. arxiv.org/abs/2411.04996 (1/n)
Weixin Liang
@liang_weixin
Nov 8, 2024
How can we reduce pretraining costs for multi-modal models without sacrificing quality? We study this Q in our new work: arxiv.org/abs/2411.04996 At @AIatMeta, We introduce Mixture-of-Transformers (MoT), a sparse architecture with modality-aware sparsity for every non-embedding
32K
Victoria X Lin
@VictoriaLinML
Jul 6, 2020
Excited to introduce Photon 🌌, a deep-learning powered cross-domain natural language interface to data systems. Try it out at the link below. #acl2020nlp Live demo: naturalsql.com Paper: aclweb.org/anthology/2020… ACL conference virtual talk: virtual.acl2020.org/paper_demo.130…
Victoria X Lin
@VictoriaLinML
Oct 22, 2021
Our team @facebookai is hiring NLP research interns. If you're pursuing a PhD and would like to spend 16w solving impactful NLP problems, please consider applying. You can DM me your CV or directly apply at facebookcareers.com/jobs/111678068…
Victoria X Lin
@VictoriaLinML
Aug 25, 2020
Talk to your databases -- Photon 🌌 v1.1 is here! What's new? - A dual input mode that accepts both NL and SQL queries - Enhanced GUI with more test databases - New blog post Blog post: blog.einstein.ai/talk-to-your-d… Live demo: naturalsql.com ArXiv: arxiv.org/abs/2007.15280
GIF
Victoria X Lin
@VictoriaLinML
Dec 22, 2022
💫 It was a fun couple of months working on instruction fine-tuning language models at large scale. The enormous capacity of LLMs has the potential to shift traditional task-specific modeling to developing generalist models that can perform a wide and growing range of tasks. 1/n
AI at Meta
@AIatMeta
Dec 22, 2022
Announcing OPT-IML: a new language model from Meta AI with 175B parameters, fine-tuned on 2,000 language tasks — openly available soon under a noncommercial license for research use cases. Research paper & more details on GitHub ⬇️
59K
Victoria X Lin
@VictoriaLinML
Dec 24, 2020
Mapping natural language to complex SQL queries across different databases demands modeling of both the DB relations and SQL syntax structures, so how well can Seq2Seq style models perform on such tasks? ❄️ Paper: arxiv.org/pdf/2012.12627… ❄️ Code: github.com/salesforce/Tab…
Victoria X Lin
@VictoriaLinML
Nov 16, 2019
🔨I made this #Python tool which converts a list of paper titles to their DBLP #BibTex entries saved in a .bib file. Check it out: github.com/todpole3/accio… Powered by the surprisingly easy-to-use DBLP API: dblp.uni-trier.de/faq/13501473
GitHub - todpole3/accio-bib: Read a list of paper titles and save their DBLP bibtex entries in a...
From github.com
Victoria X Lin
@VictoriaLinML
Dec 21, 2021
Excited to share our work @MetaAI on large-scale multilingual autoregressive language modeling. We present new SoTA on few-shot multilingual commonsense reasoning, NLI and machine translation on a large number of language pairs. Paper, model and code released👇
Xian Li
@xl_nlp
Dec 21, 2021
🌍Few-shot learning beyond English🌏 📢 Announcing XGLMs, a series of multilingual autoregressive languages models setting new SoTA on few-shot learning and outperforming English-centric models (e.g. GPT-3). Paper: arxiv.org/abs/2112.10668 Models and code: github.com/pytorch/fairse…
Victoria X Lin
@VictoriaLinML
Sep 3, 2018
QA over incomplete knowledge base as graph search. In our #emnlp2018 paper arxiv.org/abs/1808.10568, we highlighted issues with on-policy RL caused by false negative predictions and false positive search paths and proposed solutions. Joint work w/ @CaimingXiong and @RichardSocher
Victoria X Lin
@VictoriaLinML
Jul 28, 2023
I'm delighted to give an invited talk at the KLR workshop at #ICML2023 about LLMs and instruction following. Starting at 2pm HST in Room 301. Looking forward to the discussion and exchange.
Nezihe Merve Gürel (nmervegurel.bsky.social)
@nmervegurel
Jul 25, 2023
So excited for our upcoming Knowledge and Logical Reasoning Workshop at @icmlconf Join us on Friday, July 28th in Room 301 to see many many wonderful talks⤵️ and posters: klr-icml2023.github.io/papers.html. Full schedule here: klr-icml2023.github.io/schedule.html
7.6K
Victoria X Lin
@VictoriaLinML
Dec 12, 2024
#NeurIPS2024 I will present "Nearest Neighbor Speculative Decoding for LLM Generation and Attribution" led by @alexlimh23 at the poster session today. ⏰ Thu Dec 12 at 4:30-7:30 PM PST 🏛️ East Exhibit Hall A-C, #2201 🔗 neurips.cc/virtual/2024/p… Please drop by if you would like to
Minghan
@alexlimh23
Dec 6, 2024
1/ Excited to share that our paper "NEST🪺: Nearest Neighbor Speculative Decoding for LLM Generation and Attribution" is accepted at #NeurIPS2024! 🚀 Catch us at the poster session on Thu, Dec 12, 4:30–7:30 PM PST, East Exhibit Hall A-C, #2201. [Details: neurips.cc/virtual/2024/p…]
8.2K