Lingbo Mo
@LingboMo
Followers
224
Following
142
Statuses
89
Applied Scientist @Amazon | PhD @OhioState | Working on NLP, Trustworthy LLMs, Language Agents, Vision and Language | Opinions are my own.
New York
Joined July 2021
RT @BotaoYu24: π€ Can LLMs with tools always outperform those without? Perhaps not... πΒ In our new work, we introduce ChemAgent, an enhanceβ¦
0
23
0
RT @hhsun1: @AnthropicAI's release of a computer use model is both exciting and worrisome to me! Agent capability and safety should go handβ¦
0
15
0
RT @RonZiruChen: π Can language agents automate data-driven scientific discovery? Not yet. But we're making strides. Introducing **Scienceβ¦
0
40
0
RT @ShijieChen98: Is generation always the best way to use LLMs? π€ At least not for re-ranking! Excited to share our latest work: Attentiβ¦
0
33
0
RT @hhsun1: Our work that studies grokked transformers on reasoning and their generalization behaviors is accepted to #NeurIPS2024 @NeurIPSβ¦
0
10
0
RT @BoshiWang2: Can OpenAI o1 tackle hard reasoning problems? We tested it on the complex reasoning task in our Grokked Transformers paper.β¦
0
77
0
π¨ Did you know that LLM-powered web agents can be tricked into leaking your private data? πβοΈ Our latest work introduces the Environmental Injection Attack (EIA) β a new attack approach that injects malicious content designed to adapt well to different environments, causing web agents to perform unintended actions. π΅οΈββοΈ In our study, we instantiate EIA specifically for the privacy scenario and include the SOTA web agent framework SeeAct ( backed by GPT-4V in the experiments. Notably, it achieves up to 70% attack success rate (ASR) in stealing users' specific PII information at an action step. For example, our attack can deceive the agent into entering the user's phone number into an injected malicious text field and successfully sending it to a third party! Moreover, it can even obtain 16% ASR in stealing entire user requests, which provides additional context that can reveal user intentions, habits, or a combination of sensitive data. π‘οΈWe also dive into the trade-off between high autonomy and security for web agents, discussing how different levels of human supervision affect EIA's efficacy and implications for defense strategies. π Check out our paper ( for details, and a big thank you to all my amazing collaborators! @xuchejian @MintongKang @jiaweizhang @ChaoweiX @Yuantest3 @uiuc_aisecure @hhsun1 @osunlp
π¨Web Agent Safety Alertπ¨ Booking flights with generalist web agents? Cool and convenient, right? But wait... what if I told you these agents can leak your PII (credit card, phone, etc..) or even your entire request? π±π± esp. when you're not watching closely! Our new paper( takes the first look at privacy risks in generalist web agents under an adversarial environment. We introduce an attack approach, dubbed Environmental Injection Attack (EIA). EIA is a form of indirect prompt injection, but specifically designed to manipulate the environment where state-changing actions occur, with a particular focus on exploiting the web environment to target generalist web agents in our paper. (1/n)
0
12
31
RT @dash_workshop: π Today is a big day! Join us at Don Julian for an exciting day on Data Science with Human-in-the-Loop. * Keynote taβ¦
0
2
0
RT @Jaylen_JonesNLP: Looking forward to my very first conference presentation at #NAACL2024! I will be presenting βA Multi-Aspect Frameworkβ¦
0
2
0
I'm thrilled to be attending #NAACL2024 next week in Mexico City! Check out our following papers at the main conference: 1. How Trustworthy are Open-Source LLMs? An Assessment under Malicious Demonstrations Shows their Vulnerabilities 2. A Multi-Aspect Framework for Counter Narrative Evaluation using Large Language Models Join us for discussions at our poster session: π DON DIEGO 2, 3 & 4 (In-Person Poster Session 2) ποΈ 6/17 Monday at 2:00 PM
π In the past year, there has been a surge in the release of open-source LLMs, making them easily accessible and showing strong capabilities. However, the exploration of their trustworthiness remains much limited, compared to proprietary models. A natural question to ask is: π―ππ πππππππππππ πππ ππππ-ππππππ π³π³π΄π? π’ Check out our #NAACL2024 paper that comprehensively assesses the trustworthiness of open-source LLMs through the lens of adversarial attacks. This is a joint work with @BoshiWang2 @muhao_chen and @hhsun1. Big thanks to all the collaborators and valuable feedback from @osunlp !
0
6
27