INDEX

Explanations

proper nouns or names associated with legal or police contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 37-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_37/width_16k/average_l0_124

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.37.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

@"/

-0.38

 ddelweddau

-0.36

лтемелер

-0.36

 cuno

-0.36

 jsonify

-0.34

 Italijanski

-0.33

 AssemblyTitle

-0.33

PasswordEncoder

-0.33

Hochspringen

-0.32

flashdata

-0.32

POSITIVE LOGITS

 standing

1.34

 sitting

1.20

 holding

1.10

 seated

1.07

 dressed

1.02

 kneeling

1.02

 walking

1.00

 lounging

0.98

 wearing

0.98

 Standing

0.98

Activations Density 0.412%

proper nouns or names associated with legal or police contexts

No Comments

No Known Activations