Interpretability — full archive
11 daily briefings · 2 reviews.
$ ls interpretability/reviews/ # the step-back reads
- week TransformerLens 4.2.0 fixes Cohere bridge and optimizer parity
- week SAELens 6.53 ships bidirectional AbsTopK and whitening
- nnsight 0.8.0 cuts import time from 7.6s to 1.6s
- nnsight 0.8.0 cuts import time, adds a left-padding fix
- SAELens 6.54.5 fixes silent feature-density skips
- Gemma Scope 2 affine transcoders loaded wrongly, SAELens sparsity stats off by one
- TransformerLens 4.2.0 fixes Cohere bridge and optimizer parity
- All quiet as TransformerLens trims slow tests
- Slow PGD ladders read out from one attack run
- SAELens 6.54.0 folds decoder norms at load
- SAELens 6.53.1 keeps excluded tokens untouched during evals
- SAELens v6.53.0 adds bidirectional AbsTopK; param-decomp expands LM targets
- SAELens fixes a training deadlock and a JumpReLU load failure
Never miss a Interpretability briefing.
Get Interpretability's week in review every Monday - and security advisories the day they land. Free, unsubscribe in one click.
One click. Every Monday: what mattered across the projects you've been reading (or our AI picks) — security news same-day. You refine anytime. Choose specific journals · Read a sample issue →