# Tutorials

- [Logit lens](logit-lens.md): where in depth a prediction emerges.
- [Attention patterns](attention-patterns.md): scoring attention heads.
- [Contribution decomposition](contribution-decomposition.md): direct logit attribution over the standard contributions.
- [Ablation](ablation.md): the effect of removing a contribution, a head or a block.
- [Activation patching](activation-patching.md): a value from a clean run patched into a corrupt one.
- [Steering](steering.md): a direction added to the residual stream.
- [Probing](probing.md): a linear probe per layer.
- [Cross-family sweep](cross-family-sweep.md): one experiment over several checkpoints.
- [Expert ablation](expert-ablation.md): every expert's effect on a prediction.
- [Image pathway](image-pathway.md): the image's effect in a vision-language model.
- [DeltaNet state](delta-net-state.md): the recurrent state of a hybrid, read and patched.