Tom Lieberum

Master Student in AI at University of Amsterdam

I wrote a small library for transformer interpretability:

Right now I'm really interested in grokking and mechanistic interpretability more generally.

Wiki Contributions


DeepMind is hiring for the Scalable Alignment and Alignment Teams

I can't speak to the option for remote work but as a counterpoint, it seems very straightforward to get a UK visa for you and your spouse/children (at least straightforward relative to the US). The relevant visa to google is the Skilled Worker / Tier 2 visa if you want to know more.

ETA: Of course, there are still legitimate reasons for not wanting to move. Just wanted to point out that the legal barrier is lower than you might think.

Hoagy's Shortform

There is definitely something out there, just can't recall the name. A keyword you might want to look for is "disentangled representations".

One start would be the beta-VAE paper

Replacing Karma with Good Heart Tokens (Worth $1!)

Considering you get at least one free upvote from posting/commenting itself, you just have to be faster than the downvoters to generate money :P

Gears-Level Mental Models of Transformer Interpretability

Small nitpick:

The PCA plot is using the smallest version of GPT2, and not the 1.5B parameter model (that would be GPT2-XL). The small model is significantly worse than the large one and so I would be hesitant to draw conclusions from that experiment alone.

Do a cost-benefit analysis of your technology usage

I want to second your first point. Texting frequently with significant others lets me feel be part of their life and vice versa which a weekly call does not accomplish, partly because it is weekly and partly because I am pretty averse to calls. 

In one relationship I had, this led to significant misery on my part because my partner was pretty strict on their phone usage, batching messages for the mornings and evenings. For my current primary relationship, I'm convinced that the frequent texting is what kept it alive while being long-distance. 

To reconcile the two viewpoints, I think it is still true that superficial relationships via social media likes or retweets are not worth that much if they are all there is to the relationship. But direct text messages are a significant improvement on that. 

Re your blog post:
Maybe that's me being introverted but there are probably significant differences in whether people feel comfortable/like texting or calling. For me, the instantaneousness of calling makes it much more stressful, and I do have a problem with people generalizing either way that one way to interact over distances is superior in general. I do cede the point that calling is of course much higher bandwidth, but it also requires more time commitment and coordination. 

Hypothesis: gradient descent prefers general circuits

I tried increasing weight decay and increased batch sizes but so far no real success compared to 5x lr. Not going to investigate this further atm.

Hypothesis: gradient descent prefers general circuits

Oh I thought figure 1 was S5 but it actually is modular division. I'll give that a go..

Here are results for modular division. Not super sure what to make of them. Small increases in learning rate work, but so does just choosing a larger learning rate from the beginning. In fact, increasing lr to 5x from the beginning works super well but switching to 5x once grokking arguably starts just destroys any progress. 10x lr from the start does not work (nor when switching later)

So maybe the initial observation is more a general/global property of the loss landscape for the task and not of the particular region during grokking?

Hypothesis: gradient descent prefers general circuits

So I ran some experiments for the permutation group S_5 with the task x o y = ?

Interestingly here increasing the learning rate just never works. I'm very confused.

Hypothesis: gradient descent prefers general circuits

I updated the report with the training curves. Under default settings, 100% training accuracy is reached after 500 steps.

There is actually an overlap between the train/val curves going up. Might be an artifact of the simplicity of the task or that I didn't properly split the dataset (e.g. x+y being in train and y+x being in val). I might run it again for a harder task to verify.

Hypothesis: gradient descent prefers general circuits

Yep I used my own re-implementation, which somehow has slightly different behavior.

I'll also note that the task in the report is modular addition while figure 1 from the paper (the one with the red and green lines for train/val) is the significantly harder permutation group task.

Load More