RecNet
AboutHelpAll Users
Language Models with Conformal Factuality GuaranteesChristopher Mohri, Tatsunori Hashimoto
2024
Read
Tatsunori Hashimoto recommended on on 2/27/2024
Absolutely and completely shameless self-promotion, but it's got a very fun idea (LM outputs are actually also sets) and uses a very cool set of tools (conformal prediction) so hopefully it's worth a few minute skim.
Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion TokensJiacheng Liu, Sewon Min, Luke Zettlemoyer, Yejin Choi, Hannaneh Hajishirzi
2024
Read
Tatsunori Hashimoto recommended on on 2/6/2024
N-grams with 1 trillion tokens! Fast suffix arrays! What's not to like? (They don't compute perplexities or generate from the ngram model.. but otherwise, super cool thing to try!)
Paloma: A Benchmark for Evaluating Language Model FitIan Magnusson, Akshita Bhagia, Valentin Hofmann, Luca Soldaini, Ananya Harsh Jha, Oyvind Tafjord, Dustin Schwenk, Evan Pete Walsh, Yanai Elazar, Kyle Lo, Dirk Groeneveld, Iz Beltagy, Hannaneh Hajishirzi, Noah A. Smith, Kyle Richardson, Jesse Dodge
2023
Read
Tatsunori Hashimoto recommended on on 1/23/2024
This is a pretty interesting multi-domain perplexity eval. Especially the multi-domain evals seem like a useful resource for thinking about 'are there tradeoffs between different domain perplexities', and 'whats the right perplexity mix to optimize for?'
Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in MedicineHarsha Nori, Yin Tat Lee, Sheng Zhang, Dean Carignan, Richard Edgar, Nicolo Fusi, Nicholas King, Jonathan Larson, Yuanzhi Li, Weishung Liu, Renqian Luo, Scott Mayer McKinney, Robert Osazuwa Ness, Hoifung Poon, Tao Qin, Naoto Usuyama, Chris White, Eric Horvitz
2023
Read
Tatsunori Hashimoto recommended on on 12/12/2023
Once again, in-context learning emerges as a distressingly strong baseline.
Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2Hamish Ivison, Yizhong Wang, Valentina Pyatkin, Nathan Lambert, Matthew Peters, Pradeep Dasigi, Joel Jang, David Wadden, Noah A. Smith, Iz Beltagy, Hannaneh Hajishirzi
2023
Read
Tatsunori Hashimoto recommended on on 11/28/2023
Cool new paper from AI2 on studying impacts of instruction-tuning data, and trying to push the performance of fine-tuning LMs.
© 2024 RecNet. All rights reserved.