Are you into generative AI for document retrieval, grounded answer generation, recommendation, knowledge graphs and other modalities?
Our paper submission deadline is coming up:
🗓 April 25th
Gen-IR #SIGIR2024
coda.io/@sigir/gen-ir-…

🚨 New paper with Yang Gao (Yang Gao), Dana Alon and Don Metzler (@metzlerd).

🧪 Best-of-Venom: Attacking RLHF by Injecting Poisoned Preference Data

📝 arxiv.org/abs/2404.05530

We demonstrate a data poisoning attack in RLHF that manipulates the LM generations. A thread 🧵

Ruqing Zhang, Don Metzler, Andrew Yates and Ziyan Jiang are proud to announce our Call for Papers for the 2nd “Generative Information Retrieval Workshop” #SIGIR2024 🎉

🗓 deadline: April 25

👉 CfP: coda.io/@sigir/gen-ir-…

#GenerativeIR #GenerativeAI

The submission deadline for the TOIS Special Issue on Pre-trained Models for Search and Recommendation is approaching! Don't miss the chance to contribute your insights and research on pre-trained models for IR! #TOIS dl.acm.org/pb-assets/stat…

Happy that our report on the GenIR workshop in the SIGIR Forum Dec 2023 is out!

We collected #rag hot takes🔥and research ideas 🧪 in the last pages:
sigir.org/wp-content/upl…

coauthors: Ruqing Zhang, Don Metzler, Andrew Yates, Romain Deffayet, Philipp Hager and Sami Jullien

Please be sure to find Vinh Q. Tran Sanket Vaibhav Mehta, Ph.D. Kai Hui and/or Ronak Pradeep if you're at EMNLP and want to learn more about generative IR and/or Differentiable Search Indexes. I hear you might also find Yi Tay lurking in dark corners near the venue.

New paper: 'SEMQA: Semi-Extractive Multi-Source Question Answering'

Answers combine extracted spans with free text -> better attributions

Amazing colab w. Adam L. Haitian Sun
Jai Gupta Jonathan Berant William Cohen Don Metzler (@GoogleAI,Google DeepMind)

arxiv.org/abs/2311.04886

✨ New Paper ✨

Deep dive on demonstrations to enhance LLM-based passage ranking 🚀 insights for pointwise ranking using query likelihood 🚀

huggingface.co/papers/2310.14…

👏Call For TOIS Special Section on Pre-trained Models for Search and Recommendation
Submissions deadline: Feb 1, 2024

Excited for tomorrow's Gen-IR workshop SIGIR 2024 #SIGIR2023 !
Three panels, a poster session and a roundtable discussion!
ROOM: North Hall 3F – 9 am
Andrew Yates Don Metzler Jiafeng Guo Vinh Tran Rajhans Samdani Chiirag Shah @[email protected] on Mastodon Guido Zuccon Omar Khattab Fabio Petroni

congratulations, Lidan, Don Metzler, and Jimmy Lin!

Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting

Proposes PRP, which uses the query and a pair of documents as the prompt for LLMs to perform ranking tasks.

arxiv.org/abs/2306.17563

We just officialised the list of 15 accepted papers on our #GenIR at #SIGIR2023 website 👏:
coda.io/@sigir/gen-ir

PDFs coming up soon..

It’s been slightly more than a year since the UL2 paper (arxiv.org/abs/2205.05131) was released.

Here’s a summary thread of some notable models/research papers that use the UL2 objective for training (aside from the original UL2/Flan-UL2 of course).

🧵 thread below

#1 -

Led by our student researcher Ronak Pradeep, our new preprint answers the #1 question we receive about DSI & Generative Retrieval, and represents what we currently understand about the paradigm. IMO it's a must read for those looking to do research or application in this area. 1/n

We're extending the submission deadline for the Gen-IR workshop [coda.io/@sigir/gen-ir] to May 2 so that there is a little bit of time between SIGIR camera-ready and our workshop 😀

Happy to share our recent work 'Recommender Systems with Generative Retrieval'!

Joint work with Shashank Rajput, Nikhil Mehta, Yi Tay, Vinh Q. Tran and other awesome colleagues at Google Brain, Research, and YouTube.

Preprint: shashankrajput.github.io/Generative.pdf

#GenerativeAI
🧵 (1/n)

Listening to Don Metzler (Google Research) talking about Generative Information Retrieval at the UofG Computing Science Glasgow IR Group talk series #IRTalk

Very excited to be co-organizing the 1st Generative Information Retrieval Workshop (at SIGIR '23) with Gabriel Bénédict and Ruqing Zhang. Please check out the CfP and consider submitting relevant papers to the workshop!

New open source Flan-UL2 20B checkpoints :)
- Truly open source 😎 No forms! 🤭 Apache license 🔥
- Best OS model on MMLU/Big-Bench hard 🤩
- Better than Flan-T5 XXL & competitive to Flan-PaLM 62B.
- Size ceiling of Flan family just got higher!

Blog: yitay.net/blog/flan-ul2-…