2022

Mass-Editing Memory in a Transformer

Meng, Kevin, Sharma, Arnab Sen, Andonian, Alex et al.

Understand

Recent work has shown exciting promise in updating large language models with new memories, so as to replace obsolete information or add specialized knowledge.

However, this line of work is predominantly limited to updating single associations.
We develop MEMIT, a method for directly updating a language model with many memories, demonstrating experimentally that it can scale up to thousands of associations for GPT-J (6B) and GPT-NeoX (20B), exceeding prior work by orders of magnitude.
Our code and data are at https://memit.baulab.info.

Built on

Preprogramming for mechanical translation
Richard H Richens · 1956
Earlier work this paper cites.
A simple neural network generating an interactive memory
James A Anderson · 1972
Earlier work this paper cites.
Correlation matrix memories
Teuvo Kohonen · 1972
Earlier work this paper cites.
A framework for representing knowledge, 1974
Marvin Minsky · 1974
Earlier work this paper cites.
What is a knowledge representation?
Randall Davis, Howard Shrobe, and Peter Szolovits · 1993
Earlier work this paper cites.
Introduction to linear algebra
Gilbert Strang · 1993
Earlier work this paper cites.
Cyc: A large-scale investment in knowledge infrastructure
Douglas B Lenat · 1995
Earlier work this paper cites.
Wordnet: a lexical database for english
George A Miller · 1995
Earlier work this paper cites.
Direct and indirect effects
Judea Pearl · 2001
Earlier work this paper cites.
Dbpedia: A nucleus for a web of open data
Sören Auer, Christian Bizer, Georgi Kobilarov, Jens Lehmann, Richard Cyganiak, and Zachary Ives · 2007
Earlier work this paper cites.
Freebase: A shared database of structured general human knowledge
Kurt Bollacker, Robert Cook, and Patrick Tufts · 2007
Earlier work this paper cites.
Conceptnet: A lexical resource for common sense knowledge
Catherine Havasi, Robert Speer, and Jason Alonso · 2007
Earlier work this paper cites.
Yago: a core of semantic knowledge
Fabian M Suchanek, Gjergji Kasneci, and Gerhard Weikum · 2007
Earlier work this paper cites.
Toward an architecture for never-ending language learning
Andrew Carlson, Justin Betteridge, Bryan Kisiel, Burr Settles, Estevam R Hruschka, and Tom M Mitchell · 2010
Earlier work this paper cites.
Knowledge vault: A web-scale approach to probabilistic knowledge fusion
Xin Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Murphy, Thomas Strohmann, Shaohua Sun, and Wei Zhang · 2014
Earlier work this paper cites.
Wikidata: a free collaborative knowledgebase
Denny Vrandečić and Markus Krötzsch · 2014
Earlier work this paper cites.
Zero-shot relation extraction via reading comprehension
Omer Levy, Minjoon Seo, Eunsol Choi, and Luke Zettlemoyer · 2017
Earlier work this paper cites.
Attention is all you need
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin · 2017
Earlier work this paper cites.

Similar

Comet: Commonsense transformers for automatic knowledge graph construction
Antoine Bosselut, Hannah Rashkin, Maarten Sap, Chaitanya Malaviya, Asli Celikyilmaz, and Yejin Choi · 2019
Cited alongside, same era.
Pytorch: An imperative style, high-performance deep learning library
Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al · 2019
Cited alongside, same era.
Language models as knowledge bases?
Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller · 2019
Cited alongside, same era.
Language models are unsupervised multitask learners
Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al · 2019
Cited alongside, same era.
Editable neural networks
Anton Sinitsin, Vsevolod Plokhotnyuk, Dmitry Pyrkin, Sergei Popov, and Artem Babenko · 2019
Cited alongside, same era.
Huggingface’s transformers: State-of-the-art natural language processing
Original
Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, et al · 2019
Cited alongside, same era.
Language models are few-shot learners
Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, and Dario Amodei · 2020
Cited alongside, same era.
How can we know what language models know?
Zhengbao Jiang, Frank F Xu, Jun Araki, and Graham Neubig · 2020
Cited alongside, same era.
How context affects language models’ factual predictions
Fabio Petroni, Patrick Lewis, Aleksandra Piktus, Tim Rocktäschel, Yuxiang Wu, Alexander H Miller, and Sebastian Riedel · 2020
Cited alongside, same era.
How much knowledge can you pack into the parameters of a language model?
Adam Roberts, Colin Raffel, and Noam Shazeer · 2020
Cited alongside, same era.
Autoprompt: Eliciting knowledge from language models with automatically generated prompts
Taylor Shin, Yasaman Razeghi, Robert L Logan IV, Eric Wallace, and Sameer Singh · 2020
Cited alongside, same era.
Investigating gender bias in language models using causal mediation analysis
Jesse Vig, Sebastian Gehrmann, Yonatan Belinkov, Sharon Qian, Daniel Nevo, Yaron Singer, and Stuart M Shieber · 2020
Cited alongside, same era.

Then

Mind the gap: Assessing temporal generalization in neural language models
Angeliki Lazaridou, Adhi Kuncoro, Elena Gribovskaya, Devang Agrawal, Adam Liska, Tayfun Terzi, Mai Gimenez, Cyprien de Masson d’Autume, Tomas Kocisky, Sebastian Ruder, et al · 2021
Later among the works it cites.
Fast model editing at scale, 2021
Eric Mitchell, Charles Lin, Antoine Bosselut, Chelsea Finn, and Christopher D. Manning · 2021
Later among the works it cites.
Carbon emissions and large neural network training
Original
David Patterson, Joseph Gonzalez, Quoc Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David So, Maud Texier, and Jeff Dean · 2021
Later among the works it cites.
Relational world knowledge representation in contextual language models: A review
Tara Safavi and Danai Koutra · 2021
Later among the works it cites.
GPT-J-6B: A 6 Billion Parameter Autoregressive Language Model
Ben Wang and Aran Komatsuzaki · 2021
Later among the works it cites.
Knowledge graphs 2021: a data odyssey
Gerhard Weikum · 2021
Later among the works it cites.
Temporal effects on pre-trained models for language processing tasks
Oshin Agarwal and Ani Nenkova · 2022
Closest in time.
A review on language models as knowledge bases
Original
Badr AlKhamissi, Millicent Li, Asli Celikyilmaz, Mona Diab, and Marjan Ghazvininejad · 2022
Closest in time.
Gpt-neox-20b: An open-source autoregressive language model, 2022
Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, and Samuel Weinbach · 2022
Closest in time.
Palm: Scaling language modeling with pathways
Original
Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, et al · 2022
Closest in time.
Knowledge neurons in pretrained transformers
Damai Dai, Li Dong, Yaru Hao, Zhifang Sui, Baobao Chang, and Furu Wei · 2022
Closest in time.
Analyzing transformers in embedding space
Original
Guy Dar, Mor Geva, Ankit Gupta, and Jonathan Berant · 2022
Closest in time.
Transformer feed-forward layers build predictions by promoting concepts in the vocabulary space
Original
Mor Geva, Avi Caciularu, Kevin Ro Wang, and Yoav Goldberg · 2022
Closest in time.
StreamingQA: A benchmark for adaptation to new knowledge over time in question answering models
Adam Liska, Tomas Kocisky, Elena Gribovskaya, Tayfun Terzi, Eren Sezener, Devang Agrawal, D’Autume Cyprien De Masson, Tim Scholtes, Manzil Zaheer, Susannah Young, et al · 2022
Closest in time.
Locating and editing factual associations in GPT
Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov · 2022
Closest in time.
Memory-based model editing at scale
Eric Mitchell, Charles Lin, Antoine Bosselut, Chelsea Finn, and Christopher D. Manning · 2022
Closest in time.
Kformer: Knowledge injection in transformer feed-forward layers
Yunzhi Yao, Shaohan Huang, Li Dong, Furu Wei, Huajun Chen, and Ningyu Zhang · 2022
Closest in time.
Modifying memories in transformer models, 2020
Original
Chen Zhu, Ankit Singh Rawat, Manzil Zaheer, Srinadh Bhojanapalli, Daliang Li, Felix Yu, and Sanjiv Kumar · 2022
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…