INDEX

Explanations

terms associated with learned associations in attention and stimulus processing

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-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

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

ſelf

-0.66

MemoryWarning

-0.64

 purpoſe

-0.62

AddTagHelper

-0.62

 houſe

-0.61

 AssemblyTitle

-0.61

 Chriftian

-0.60

 Houſe

-0.59

 Majefty

-0.58

AccessorTable

-0.57

POSITIVE LOGITS

 flash

0.33

 repeatedly

0.32

 melodies

0.31

 manipulated

0.30

 constantemente

0.30

 diha

0.30

 naar

0.29

 cues

0.29

teils

0.28

 susc

0.28

Activations Density 0.111%

terms associated with learned associations in attention and stimulus processing

No Comments

No Known Activations