INDEX

Explanations

numbers indicating percentage, particularly '90'

oai_token-act-pair · gpt-3.5-turbo

references to the percentage "90" within various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

lay

-0.78

board

-0.77

holder

-0.74

folk

-0.73

REDACTED

-0.73

holding

-0.72

hold

-0.71

wan

-0.71

maid

-0.70

stand

-0.69

POSITIVE LOGITS

%:

1.00

0.98

%"

0.96

idth

0.96

 ILCS

0.96

Â°

0.93

orsi

0.90

Âº

0.87

ength

0.87

 degrees

0.85

Activations Density 0.027%

numbers indicating percentage, particularly '90'

references to the percentage "90" within various contexts

No Comments

No Known Activations