INDEX

Explanations

names and titles of individuals in leadership or administrative roles

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 36-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_36/width_16k/average_l0_120

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.36.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Head

-0.65

 Heads

-0.54

 Guest

-0.51

 Directors

-0.48

 Main

-0.47

 Season

-0.47

Head

-0.46

 Coach

-0.46

 Headquarters

-0.45

 Patron

-0.45

POSITIVE LOGITS

 spokesman

0.71

 spokeswoman

0.63

 chairman

0.61

MigrationBuilder

0.60

 sociology

0.60

 executive

0.60

 director

0.60

 chair

0.59

 spokesperson

0.58

 deputy

0.58

Activations Density 0.287%

names and titles of individuals in leadership or administrative roles

No Comments

No Known Activations