INDEX

Explanations

themes related to power dynamics and control

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_16k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Stubbs

-0.43

 trùng

-0.41

/*

-0.40

listItem

-0.38

ECM

-0.37

 conclusão

-0.36

 Seen

-0.34

 Crist

-0.33

싶

-0.33

 Absch

-0.33

POSITIVE LOGITS

 Power

1.48

 power

1.45

Power

1.44

power

1.41

 POWER

1.38

POWER

1.37

 Powers

1.16

Powers

1.14

 powers

1.10

setPower

1.09

Activations Density 0.120%

themes related to power dynamics and control

No Comments

No Known Activations