INDEX

Explanations

phrases related to effort, resources, and work

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

having

-0.56

 Rely

-0.54

 Having

-0.54

fromnode

-0.53

 having

-0.53

 AssemblyCulture

-0.51

Having

-0.49

 recurrir

-0.49

 relying

-0.49

 Spending

-0.48

POSITIVE LOGITS

 solving

1.13

 repairing

1.04

 improving

1.02

 resolving

1.02

 restoring

1.01

 diagnosing

1.00

 perfecting

0.93

 rescuing

0.93

solving

0.92

 educating

0.91

Activations Density 1.129%

phrases related to effort, resources, and work

No Comments

No Known Activations