INDEX

Explanations

adjectives or verbs indicating persuasiveness or believability

oai_token-act-pair · gpt-3.5-turbo

terms related to persuasion and convincing arguments

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

dule

-0.69

heid

-0.69

encies

-0.68

ews

-0.67

 Services

-0.67

ycle

-0.66

lus

-0.63

 Arts

-0.63

tions

-0.63

tion

-0.62

POSITIVE LOGITS

 convincing

3.88

 persu

2.07

 convinc

2.06

 persuasive

1.95

 compelling

1.84

 believable

1.79

 reassuring

1.51

 convince

1.50

 plausible

1.46

 convin

1.39

Activations Density 0.015%

adjectives or verbs indicating persuasiveness or believability

terms related to persuasion and convincing arguments

No Comments

No Known Activations