INDEX

Explanations

terms associated with hunting and feeding behaviors

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_33/width_131k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.33.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

PullParser

-0.45

atorship

-0.42

 propeller

-0.38

 poussière

-0.37

miral

-0.36

 beso

-0.36

ixels

-0.36

 getApplication

-0.36

ymbol

-0.35

violation

-0.35

POSITIVE LOGITS

 prey

1.29

prey

0.96

 Prey

0.86

 victims

0.73

 victim

0.67

 fish

0.66

 pray

0.65

 mice

0.65

猎

0.63

 Victims

0.58

Activations Density 0.017%

terms associated with hunting and feeding behaviors

No Comments

No Known Activations