Learn how ReAct, a general paradigm for synergizing reasoning and acting in language models, presents more interpretable, diagnosable, and controllable task-solving trajectories while outperforming reasoning and acting only paradigms. Read the blog → https://
goo.gle/3TiNcDp
RESEARCH
-
ReAct: Synergizing Reasoning and Acting in Language Models
By
–
-
ROME: Efficient Factual Editing in GPT Models
By
–
MIT, Northeastern & Technion Propose ROME for Efficient Locating and Editing of Factual Associations in GPT Models
-
Exploring Large Internet Archives with Embeddings
By
–
Would you like to learn how to explore large internet archives with embeddings? Check out this co:lab friday session with our very own @JayAlammar
, sharing how he built a map of top 10,000 Hacker News posts of all time! https://
hubs.li/Q01rPlY10
#topicmodelling #embeddings -
Dr. Van Essendelft Presents Computational Physics Research on Cerebras CS-2
By
–
Join Dr. Dirk Van Essendelft at #SC22 as he describes the tremendous Computational Physics research he’s done using the Cerebras CS-2 system. There will be two presentations at 10 am and 2 pm in Cerebras booth 2833. #HPC #Physics #SC22
-
NVIDIA GPUs Power NASA Webb Telescope and AI Drug Discovery
By
–
Our GPUs interpret data from the @NASAWebb Space Telescope so that all of humanity can see outer space as never before. And to help researchers explore the vast chemical space of drug discovery, we built the #AI-powered BioNeMo cloud service. https://t.co/sf8g0w4Mda #NVIDIAstory
— NVIDIA (@nvidia) 8 novembre 2022Our GPUs interpret data from the @NASAWebb Space Telescope so that all of humanity can see outer space as never before. And to help researchers explore the vast chemical space of drug discovery, we built the #AI-powered BioNeMo cloud service. https://
nvda.ws/3xTkalW #NVIDIAstory -
Inspiration from Ajeya Cotra’s Sandwiching Concept in Research
By
–
This paper was heavily inspired by prior work, especially Ajeya Cotra's 'sandwiching' concept:
-
Human Accuracy Improves 10 Percent Through Model Interaction
By
–
Generally, we see that our participants start out with far worse accuracy on these tasks than our models do, and that by interacting with the models, they’re able do about 10 percentage points better.
-
Human-AI Collaboration Improves Task Performance Through Simple Chat Strategy
By
–
Our experiment shows that through a simple strategy – having humans chat with models while completing a task – we can help humans perform better at these tasks. This is very encouraging, albeit preliminary!
-
Non-Experts Answer Expert Questions on MMLU and QuALITY
By
–
We ask non-experts to answer expert-level questions on MMLU, and also ask people to answer questions about long QuALITY passages under a time limit that’s too short for a careful read.
-
Scalable Oversight Framework and Language Model Question-Answering Proof of Concept
By
–
Along with developing a framework for scalable oversight, we also conduct a proof of concept experiment that demonstrates a couple of question-answering tasks that work well under this paradigm with current language models: