INDEX

Explanations

instances of the word "exceed" and its variations indicating surpassing a limit or threshold

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

<strong>

-0.42

//

-0.41

Nh

-0.38

//

-0.38

 joão

-0.38

 Alger

-0.37

<em>

-0.37

 ball

-0.36

 blonde

-0.35

POSITIVE LOGITS

 exceed

1.13

 exceeded

1.04

 exceeds

1.03

 exceeding

1.03

Exceeded

0.93

 surpass

0.81

ceed

0.81

beyond

0.80

 betweenstory

0.80

 surpassing

0.77

Activations Density 0.009%

instances of the word "exceed" and its variations indicating surpassing a limit or threshold

No Comments

No Known Activations