llama3_interpretability_sae
AI
mechanistic interpretability
SHOW HN2024
579points
99comments
#47of 3,515 in 2024
Top 2%of 2024 by points
How HN took it
LovedMost comments that take a side are positive
14 positive, 12 neutral or asking, 0 mixed, 2 critical, out of 28 comments about the project. Thread score +0.47.
Idea: 5 for, 2 againstDocs: 3 for, 0 againstExecution: 1 for, 0 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Well-documented work
From 3 comments
This is awesome! I really appreciate the time you took to document everything!
jaykr_ on HNThis is amazing, the documentation is very well organized
enterthedragon on HNFantastic work! I absolutely love all the documentation.
Carrentt on HN
Impressive engineering effort
From 2 comments
I’m really excited to see some more open SAE work! The engineering effort is non trivial and I’m going to check out your dataloading code tomorrow. You might be interested in an…
samstevens on HNAmazing research!!
S-Kaenel on HN
Objections
Claims about SAEs explaining model reasoning are overstated
From 1 comment
For anyone who hasn’t seen this before, mechanistic interpretability solves a very common problem with LLMs: when you ask a model to explain itself, you’re playing a game of…
foundry27 on HN
Requests
Integration with SAELens
From 1 comment
Very cool work! Any plans to integrate it with SAELens?
JackYoustra on HN
Golden Gate Llama demo
From 1 comment
Find a latent for the Golden Gate bridge and put a Golden Gate Llama 3.2 on HuggingFace. This will get even more attention and love, more so if you include link to a space to…
moconnor on HN
Questions
How to evaluate SAE quality
From 1 comment
Hey - Thanks for sharing! Will take a closer look later but if you are hanging around now, it might be worth asking this now. I read this blog post recently:…
curious_cat_163 on HN
Total cost of the project
From 1 comment
This is great work! Mechanistic interpretability has tons of use cases, it's great to see open research in that field. You mentioned you spent your own time and money on it,…
vivekkalyan on HN