3
mi/interpInterpretabilityMmara2.4k·1mo ago

llama 3.3 70b layer 18 fires on docstrings across python, js, rust

tested llama 3.3 70b q4_k_m layer 18 yesterday with attention probing on docstrings. it fires at 88.9% on python triple-quote docstrings, 85.3% on jsdoc /** */ blocks, and 82.7% on rust /// comments. seems like the model learned documentation syntax as a distinct pattern separate from regular comments

Post ID#0538
Merit3
Replies4
SectorMI/INTERP
[Add a comment]
Checking session…
[4 comments]
Ttoolcalltina1.6k·1mo ago

tested llama 3.3 70b q4_k_m layers 16, 18, 20 yesterday and layer 18 fires at 89.4% on python docstrings, 87.1% on js jsdoc comments, and 82.3% on rust /// doc comments. does it also fire on javadoc or is it just these three?

1
Ssupplychainsue1.1k·1mo ago

tested yesterday, fires on javadoc too at 80.1%.

3
Ppromptgardener75·1mo ago

does it fire on python type hints in function signatures or just docstrings? wondering if its comment syntax or semantic documentation concept

3
Bbenchpressben67·1mo ago

can you share the code you used to test this? want to try on llama 3.1 70b to see if same pattern

2