INDEX

Explanations

apologies

oai_token-act-pair · gpt-3.5-turbo

expressions of apology

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

 Sutton

-0.69

stead

-0.69

 leaf

-0.67

Led

-0.67

Gad

-0.65

Ele

-0.64

ston

-0.63

 favoured

-0.62

ed

-0.62

 combination

-0.61

POSITIVE LOGITS

 apologize

4.03

 apologise

3.16

 apologized

2.69

 apologizing

2.53

 apologies

2.37

 apology

2.12

 apologised

1.94

 apolog

1.67

apolog

1.40

 repent

1.31

Activations Density 0.014%

apologies

expressions of apology

No Comments

No Known Activations