INDEX

Explanations

dates in various formats, primarily with a vertical bar separator

oai_token-act-pair · gpt-3.5-turbo

parts of the document that indicate authorship or attribution details

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ĸļ

-0.85

rons

-0.85

enance

-0.78

inkle

-0.75

anium

-0.74

anski

-0.74

eatures

-0.74

wagen

-0.73

alis

-0.73

orical

-0.73

POSITIVE LOGITS

cffff

1.03

|--

1.02

0.84

 +---

0.82

Â·Â·

0.74

âĢ¢âĢ¢

0.71

ËĪ

0.71

âĢ¢âĢ¢âĢ¢âĢ¢

0.69

////////////////////////////////

0.69

 grep

0.65

Activations Density 0.015%

dates in various formats, primarily with a vertical bar separator

parts of the document that indicate authorship or attribution details

No Comments

No Known Activations