INDEX

Explanations

segments of text formatted as code or data entries, including identifiers, variable names, and sequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 1-gemmascope-mlp-16k

Configuration

google/gemma-scope-2b-pt-mlp/layer_1/width_16k/average_l0_105

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.1.hook_mlp_out

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

-0.69

 itſelf

-0.52

tralight

-0.47

Its

-0.45

szer

-0.45

 againſt

-0.45

lichem

-0.45

ship

-0.44

scales

-0.43

 aurez

-0.43

POSITIVE LOGITS

1.08

0.93

0.85

SI

0.79

+:+

0.75

SI

0.74

si

0.74

 getS

0.73

ST

0.72

st

0.71

Activations Density 0.385%

segments of text formatted as code or data entries, including identifiers, variable names, and sequences

No Comments

No Known Activations