most_recent_S_attn_pat.ipynb

https://colab.research.google.com/drive/1KaqcS92-BI4FZ7m-r8rCW9tIovxA_s93#scrollTo=VcFgqbcF4YvI

Most Recent S Name Movers

Based on IOI findings, we expect to find:

• Induction heads (b/c of in-context learning)
• Name mover heads
  • Find evidence for this using Copy Scores
• Subject influencing heads (here, seems to be most recent subject)

Can be done w/ just GPT-2-small

<<<<<<

Direct Logit Attribution [IOI paper]

• Was logit diff commented on? If so, how?
  • Only states average logit difference X over Y examples. The rest of its info was only used in activation patching comparisons (no figures)
• How was activation head patching described?
  • Include the heatmap Figure

Direct Logit Attribution [IOI paper]

• average logit difference X over Y examples

<<<