INDEX

Explanations

the interjection "well" used to preface a spoken response or explanation

oai_token-act-pair · gpt-3.5-turbo

phrases that start with "Well."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 sustained

-0.65

 dash

-0.65

 visible

-0.63

 accessible

-0.62

 passing

-0.61

 identifying

-0.61

 shared

-0.61

hit

-0.61

 lasting

-0.61

 even

-0.61

POSITIVE LOGITS

Well

3.43

 Well

2.31

well

1.86

Hmm

1.61

Oh

1.60

Yeah

1.55

Anyway

1.54

Actually

1.53

Alright

1.50

Okay

1.43

Activations Density 0.013%

the interjection "well" used to preface a spoken response or explanation

phrases that start with "Well."

No Comments

No Known Activations