Sublime
An inspiration engine for ideas
Matthias Adler
@matthiasadler
Very interesting paper on using LLM generated code at Meta.
73% of recommendations passed review and is in production. WHAT. https://t.co/cna1o8BX0L
We’re looking forward to @aclmeeting #ACL2024!
From @Diyi_Yang’s group:
• How Johnny Can Persuade LLMs to Jailbreak Them
• Feedback Generation for Empowering Peer Counselors
• Silent Signals, Loud Impact: Coded Dog Whistles
• Impacts of LLM Alignment on Global Representation... See more
What We Know About Using Non-Engagement Signals in Content Ranking
arxiv.org
Who Wins Ethereum Block Building Auctions and Why?
arxiv.org
Can rerankers boost your RAG performance without a major latency tradeoff? Upon testing with @lancedb on a test set of 2k queries/ 100K docs from GooQA dataset, we saw upto 20% improvement in hit rate, with negligible latency added in most cases. More details in 🧵 https://t.co/F1AoygC5h2
We just released Mixtral 8x7B paper on Arxiv:
https://t.co/wn0kOOFTcJ https://t.co/Xth6asCjpQ