INDEX

Explanations

words related to a specific topic or subject, such as "theme" in various contexts

oai_token-act-pair · gpt-3.5-turbo

references to a central theme or motif in various contexts

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

gress

-0.71

effective

-0.70

Instant

-0.69

carry

-0.67

 sacks

-0.67

uala

-0.65

Vec

-0.63

bil

-0.63

ivably

-0.62

lessly

-0.62

POSITIVE LOGITS

 theme

3.88

 themes

2.84

 Theme

2.77

theme

2.59

Theme

2.42

 motif

1.76

 themed

1.74

 backdrop

1.36

themed

1.28

 topic

1.23

Activations Density 0.013%

words related to a specific topic or subject, such as "theme" in various contexts

references to a central theme or motif in various contexts

No Comments

No Known Activations