Kevin's picture

Kevin Meng

currently: @mit @emberml

contact: mengk [at] mit [dot] edu

location: sf or nyc probably

my interests:

  • databases
  • knowledge search
  • data visualization
  • interpretability

about me.

hi, i'm kevin! 👋 i love thinking about data, algorithms, and their impact on society.

these days, i run a small a tech company focused on wrangling and organizing knowledge. i'm also involved in efforts across mit, northeastern, and harvard to understand how deep neural nets work. i used to spend lots of time on that problem and hope to return to it someday.

aside from work & research, i care deeply about teaching. back at home, i ran a non-profit providing mentorship to aspiring scientists, where I still occasionally volunteer. at mit, i organized lots of ai workshops and taught for splash. in my free time, i enjoy long walks, cooking, reading, road trips, running very slowly, sucking at basketball, and aimlessly wandering the streets of new cities. i used to make music and miss it a lot.

i'm currently on leave from mit, where i studied computer science, electrical engineering, and a bit of math.

recent news.

  • jan 2023
    MEMIT was accepted to ICLR '23! We've scaled ROME to 100x state-of-the-art model editors [twitter] [hackernews]
  • sept 2022
    ROME will appear in NeurIPS '22! [twitter]
  • february 2022
    We've released ROME, a study on the fact-storing mechanisms in large auto-regressive transformer language models [twitter]
  • October 2020
    Our ClaimBuster model is being used by the Duke Reporter's Lab to help fact-check the 2020 Presidential Election [poynter]

things i've worked on.

Mass-Editing a Transformer Memory

Publication | Code | Project Page
Kevin Meng, Arnab Sen Sharma, Alex Andonian, Yonatan Belinkov, David Bau
*Presented at ICLR 2023 in Kigali, Rwanda

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.


Locating and Editing Factual Associations in GPT

Publication | Code | Project Page
Kevin Meng*, David Bau*, Alex Andonian, Yonatan Belinkov
*Presented at NeurIPS 2022 in New Orleans, LA

We investigate the mechanisms underlying factual knowledge recall in autoregressive transformer language models. First, we develop a causal intervention for identifying neuron activations capable of altering a model's factual predictions. Within large GPT-style models, this reveals two distinct sets of neurons that we hypothesize correspond to knowing an abstract fact and saying a concrete word, respectively. This insight inspires the development of ROME, a novel method for editing facts stored in model weights. For evaluation, we assemble CounterFact, a dataset of over twenty thousand counterfactuals and tools to facilitate sensitive measurements of knowledge editing. Using CounterFact, we confirm the distinction between saying and knowing neurons, and we find that ROME achieves state-of-the-art performance in knowledge editing compared to other methods. An interactive demo notebook, full code implementation, and the dataset are available.

VeriClaim Poster
Studying the Approximate Linearity of Apple's NeuralHash

Jagdeep Bhatia*, Kevin Meng*
*Presented at ICML 2022's ML for Cybersecurity Workshop in Baltimore, MD

Perceptual hashes map images with identical semantic content to the same n-bit hash value, while mapping semantically-different images to different hashes. These algorithms carry important applications in cybersecurity such as copyright infringement detection, content fingerprinting, and surveillance. Apple's NeuralHash is one such system that aims to detect the presence of illegal content on users' devices without compromising consumer privacy. We make the surprising discovery that NeuralHash is approximately linear, which inspires the development of novel black-box attacks that can (i) evade detection of "illegal" images, (ii) generate near-collisions, and (iii) leak information about hashed images, all without access to model parameters. These vulnerabilities pose serious threats to NeuralHash's security goals; to address them, we propose a simple fix using classical cryptographic standards.

NeuralHash Project
VeriClaim: End-to-End Computational Fact Checking

Publication | Claim-Spotter Paper | Demo Video
Kevin Meng
*Presented at NeurIPS 2021's Workshop on AI for Credible Elections

VeriClaim contains two computational modules: the claim-spotter and claim-checker. The claim-spotter first selects “check-worthy” factual statements from large amounts of text using a Bidirectional Encoder Representations from Transformers (BERT) model trained with a novel gradient-based adversarial training algorithm. Then, selected factual statements are passed to the claim-checker, which employs a separate stance detection BERT model to verify each statement using evidence retrieved from a multitude of knowledge resources. Web interface inspired by Fakta, from MIT.

VeriClaim Poster
An NLP-Powered Dashboard for Mitigating the COVID-19 Infodemic

Publication | Code | Data | Dashboard | Demo Video
Zhengyuan Zhu, Kevin Meng, Josue Caraballo, Israa Jaradat, Xiao Shi, Zeyu Zhang, Farahnaz Akrami, Haojin Liao, Fatma Arslan, Damian Jimenez, Mohammed Samiul Saeef, Paras Pathak, Chengkai Li

This paper introduces a public dashboard which, in addition to displaying case counts in an interactive map and a navigational panel, also provides some unique features not found in other places. Particularly, the dashboard uses a curated catalog of COVID-19 related facts and debunks of misinformation, and it displays the most prevalent information from the catalog among Twitter users in user-selected U.S. geographic regions. We also explore the usage of BERT models to match tweets with misinformation debunks and detect their stances. We also discuss the results of preliminary experiments on analyzing the spatio-temporal spread of misinformation.

Gradient-Based Adversarial Training on Transformer Networks

Pre-Print | Code | ClaimBuster Website
Kevin Meng*, Damian Jimenez*, Fatma Arslan, Jacob Daniel Devasier, Daniel Obembe, Chengkai Li
*Deployed on ClaimBuster, used by thousands of fact-checkers and research groups worldwide

We introduce the first adversarially-regularized, transformer-based claim spotting model that achieves state-of-the-art results by a 4.70 point F1-score margin over current approaches on the ClaimBuster Dataset. In the process, we propose a method to apply adversarial training to transformer models, which has the potential to be generalized to many similar text classification tasks. Along with our results, we are releasing our codebase and manually labeled datasets. We also showcase our models' real world usage via a live public API.

Model Architecture