Looking to safeguard your LLMs from deception?
🛡️LLM Self Defense is
🛠️Simple: No fine-tuning or preprocessing
🌐Generalizable: Works for GPT3.5 and Llama
💪Effective: Countering almost all attacks

Visit us at ICLR'24 tomorrow at Hall B #294 from 4:30pm - 6:30pm to know more!

Did you know a paper by Georgia Tech researchers received a best paper award honorable mention at ACM CHI Conference? Here is a sneak peek of Farsight as we count down to #CHI2024 ! 👀🐝

Georgia Tech Computing Georgia Tech Research
Jay Wang Polo Data Club Georgia TechAI Machine Learning at Georgia Tech

cc.gatech.edu/news/new-tool-…

Looking to convert your tables from images and PDFs to HTML? Check out our UniTable! It outperforms both GPT-4V and LLaVA-1.6 in table recognition. #TableRecognition

Paper and code in thread 🧵

#CountdowntoCHI Thanks to a new tool, developers using large language models can see potential harmful attributes in their prototypes. Farsight alerts them when they write LLM prompts that could be harmful and misused.
cc.gatech.edu/news/new-tool-…

GT@CHI: sites.gatech.edu/research/chi-2…

🤔 Want to Fine-Tune LLMs with Your Private Data for FREE? 3-in-1 tutorial designed for everyone!

📖 Accessible blog post medium.com/polo-club-of-d…
+ 💻 Step-by-step Colab notebook github.com/poloclub/Fine-…
+ 🎙️ High-level presentation

CSE research is making an impact at #CHI2024 and beyond! Here is your sneak peek on Farsight, ARCollab, and SuperNOVA, ahead of the premier international conference on human-computer interaction!

sites.gatech.edu/research/chi-2…

Polo Data Club Duen Horng "Polo" Chau Jay Wang Georgia Tech Research ACM CHI Conference

What does ML model compression have to do with HCI and design?

Over the past few years Apple we've been working in this intersection, and I’m excited to share 2 new papers on compression in practice and the importance of interactive tools.

Catch them back-to-back at #CHI2024 !

Glad to see Jay’s hard work recognized. His work does a great job of balancing the goals of making the system simple and automated (and thus easier to ship in a product) with allowing users to engage with possible harms in a reflective way.

Farsight is recognized with a #chi2024 best paper, honorable mention award!!! ❤️

Congrats to the amazing Google AI collaborators Michael Madaio, Mike Terry, Lauren Wilcox, and Chinmay Kulkarni https://hci.social/@chinmay!!!

Try our tool here ⬇️

pair-code.github.io/farsight/

For the 7th time, we’re back! The Workshop on Visualization for AI Explainability ( #visxai ) will accept “explorables” & “explainables” that visualize, explain, and tell interactive stories about AI. #ieeevis

➡️ Deadline: July 30, 2024

visxai.io

Just open-sourced this simple visualizer for ICLR 2024: github.com/kaansancak/icl….

#iclr #icml #neurips

Compare fill-in-the-blank LLM prompts to reveal learned relationships and behaviors!

🤔 Understanding what #LLMs have learned is challenging

📊 KnowledgeVIS visualizes LLM #prompts to reveal #biases and #stereotypes in text

👉 github.com/AdamCoscia/Kno…

Presenting #IEEEVIS 2024!

Visualize how LLMs grade human-written summaries with iScore!

✍️Summary writing helps students learn

🤔iScore helps engineers understand, evaluate, and trust LLMs that grade summaries before deploying in classrooms!

👉adamcoscia.com/papers/iscore/

Presenting Thur 03/21 at #IUI2024 !

Before taking it out on your roommate for leaving dirty dishes out, you probably want to practice your conflict resolution skills first. Expert conflict resolution trainers, however, are EXPENSIVE. What if we practiced with an angry LLM instead? 😈 #CHI24

arxiv.org/abs/2309.12309

Similar to how we learn digital and media literacy, we are entering an era where AI awareness is vital to understand and properly use applications.

Farsight is one such tool to help AI developers see potential issues and harm with their LLM prompts! Check it out!

Georgia Tech Computing

Can't decide between #Maui or #BigIsland w.r.t. transpo, hotels, food, beeches, etc.? Our upcoming #CHI2024 paper presents Selenite, an AI #sensemaking tool that offers comprehensive overviews, adapts to your needs, and accelerates your decision-making! arxiv.org/abs/2310.02161

Excited to share a tool Alex and I have been building!
<| 𝗲𝗻𝗱𝗼𝗳𝘁𝗲𝘅𝘁 |> is an AI-powered prompt editor that helps you write better prompts with suggested edits, automatic rewrites, and test case generation.

We'd love to get your feedback: endoftext.app

Fantastic to see this work you did at PAIR come out!

📢 Looking for current research on #HCI + #AI ? Here's a collection of 100+ #CHI2024 preprints, collected via arXiv and twitter. Suggestions welcome! dbuschek.medium.com/chi24-preprint…

A very innovative approach to envisioning RAI harms. Well done, Jay Wang!