Tiange Xiang (@xxtiange) 's Twitter Profile
Tiange Xiang

@xxtiange

CS PhD student @ Stanford @ SVL

ID: 1617915998207893505

linkhttps://ai.stanford.edu/~xtiange/ calendar_today24-01-2023 16:03:35

40 Tweet

153 Followers

40 Following

Genghan Zhang (@zhang677) 's Twitter Profile Photo

🔍 ML library development is crucial but requires expertise in ML algorithms & architecture-specific programming languages (ASPLs). đŸ€– LLM agents can enable better automation. We propose an adaptive self-improvement agentic system for generating ML libraries in STeP—a

🔍 ML library development is crucial but requires expertise in ML algorithms & architecture-specific programming languages (ASPLs).

đŸ€– LLM agents can enable better automation.  We propose an adaptive self-improvement agentic system for generating ML libraries in STeP—a
Ken Liu (@kenziyuliu) 's Twitter Profile Photo

An LLM generates an article verbatim—did it “train on” the article? It’s complicated: under n-gram definitions of train-set inclusion, LLMs can complete “unseen” texts—both after data deletion and adding “gibberish” data. Our results impact unlearning, MIAs & data transparencyđŸ§”

An LLM generates an article verbatim—did it “train on” the article?

It’s complicated: under n-gram definitions of train-set inclusion, LLMs can complete “unseen” texts—both after data deletion and adding “gibberish” data. Our results impact unlearning, MIAs & data transparencyđŸ§”
Tiange Xiang (@xxtiange) 's Twitter Profile Photo

Maybe it is not a bad idea to start a unified committee for AI conferences, which accepts rolling submissions, has only one round peer-review, and determines which venue is the best fit for a paper if accepted. What's Good is the review process will be somewhat normalized

Haoyu Xiong (@haoyu_xiong_) 's Twitter Profile Photo

Your bimanual manipulators might need a Robot Neck đŸ€–đŸŠ’ Introducing Vision in Action: Learning Active Perception from Human Demonstrations ViA learns task-specific, active perceptual strategies—such as searching, tracking, and focusing—directly from human demos, enabling robust

Agrim Gupta (@agrimgupta92) 's Twitter Profile Photo

Introducing Genie 3, our state-of-the-art world model that generates interactive worlds from text, enabling real-time interaction at 24 fps with minutes-long consistency at 720p. đŸ§”đŸ‘‡

Ken Liu (@kenziyuliu) 's Twitter Profile Photo

New paper! We explore a radical paradigm for AI evals: assessing LLMs on *unsolved* questions. Instead of contrived exams where progress ≠ value, we eval LLMs on organic, unsolved problems via reference-free LLM validation & community verification. LLMs solved ~10/500 so far:

New paper! We explore a radical paradigm for AI evals: assessing LLMs on *unsolved* questions.

Instead of contrived exams where progress ≠ value, we eval LLMs on organic, unsolved problems via reference-free LLM validation & community verification. LLMs solved ~10/500 so far:
Tiange Xiang (@xxtiange) 's Twitter Profile Photo

We are extending submission deadline to 5th Sept! Enjoy the long weekend and make sure to submit your papers in time 😎😎

Justin Johnson (@jcjohnss) 's Twitter Profile Photo

10 years ago, deep learning was in its infancy. PyTorch didn't exist. Language models were recurrent, and not large. But it felt important: a new technology that would change everything. That's why Fei-Fei Li , Andrej Karpathy, and I started CS231N Staff back in 2015 - to teach the world's