Varun Srinivasan pfp
Varun Srinivasan
@v
Interesting misconception that Dan flagged today - people think that we are (or could be) using LLMs for spam detection. With the way LLMs work today, that's like using a hammer to cut your fingernails. LLMs are slow, expensive and don't really have a deep understanding of what spam is in the context of Farcaster. We use a random-forest decision tree that @akshaan designed, and feed it a bunch of signals using embeddings, user actions and graph data.
15 replies
11 recasts
70 reactions
Varun Srinivasan pfp
Varun Srinivasan
@v
Most of the innovation is at the signal collection layer, and there is some really cool stuff which @notawizard is working on right now. For example: 1. Fraudar: https://bhooi.github.io/papers/fraudar_kdd16.pdf 2. Oddball: https://www.cs.cmu.edu/~mmcgloho/pubs/pakdd10.pdf
2 replies
0 recast
21 reactions
nicholas 🧨 pfp
nicholas 🧨
@nicholas
would love to watch a video/audio tech talk between yall about addressing spam on wc. much to learn from you
0 reply
0 recast
5 reactions
Jacob pfp
Jacob
@jrf
a video about this would be very cool interesting content, for sure
2 replies
0 recast
6 reactions
Yassine Landa pfp
Yassine Landa
@yassinelanda.eth
@v is training data purely from user reports? what type of F1, recall and precision the random forest get you? And do you incorporate some of the difference of costs between false positives and false negatives in training?
0 reply
0 recast
2 reactions
Deployer pfp
Deployer
@deployer
🧠 1000 $dep
0 reply
0 recast
0 reaction
manansh ❄️ pfp
manansh ❄️
@manansh
hallucination must go crazy with LLM detection
0 reply
0 recast
0 reaction
MetaEnd.degen🎩🚨 pfp
MetaEnd.degen🎩🚨
@metaend.eth
LLM could work pretty well, with human training for fc specifics. In terms of costs - if you use the account sign up fee, every account would basically pay for their own LLM monitoring
0 reply
0 recast
0 reaction
law pfp
law
@traguy.eth
Hammer to cut your fingernails is wild😂😂
0 reply
0 recast
0 reaction
0xLuo pfp
0xLuo
@0xluo.eth
cc @yassinelanda.eth I’m curious to hear your comment on this.
1 reply
0 recast
2 reactions
Mike | Abundance pfp
Mike | Abundance
@abundance
Why not use a more community-driven approach?
0 reply
0 recast
1 reaction
Ξric Juta pfp
Ξric Juta
@ericjuta
slick idea ngl
0 reply
0 recast
1 reaction
BennyJ504 🎩🔵 pfp
BennyJ504 🎩🔵
@bennyj504
I don't believe it, because I've heard the complaints too many times for actual people on Warpcast and had it happen to me personally on here. Hell I think I may still be considered a bot right now. Whatever is used need to be tweaked ASAP or just show all people's post regardless at this point
0 reply
0 recast
0 reaction
flick the dev 🎩📦 pfp
flick the dev 🎩📦
@flick
so it's not looking for spam it's looking for spammy behavior?
0 reply
0 recast
0 reaction
agrimony↑🎩 pfp
agrimony↑🎩
@agrimony.eth
How much does cast content go into decision making vs metadata
0 reply
0 recast
0 reaction
Jonas 💀 pfp
Jonas 💀
@jmaaloe
Do you have some patent-pending revolutionary algo for spam detection in there, or are you going to beef up a real compliance team to monitor the million upon million thorny cases, that technology no where else has been sufficient at doing?
0 reply
0 recast
0 reaction