INDEX

Explanations

instances of phrases indicating uniqueness or superlatives

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate distinct categories or classifications

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

words

-0.62

ļéĨĴ

-0.55

yards

-0.54

 carriers

-0.54

cdn

-0.53

 Minutes

-0.52

 meanwhile

-0.52

ming

-0.51

 reading

-0.50

tyard

-0.50

POSITIVE LOGITS

 kind

1.45

 type

1.11

 kinds

1.10

Kind

1.06

 sort

1.05

kind

1.02

 sorts

1.01

size

0.94

 Kind

0.93

type

0.90

Activations Density 0.079%

instances of phrases indicating uniqueness or superlatives

phrases that indicate distinct categories or classifications

No Comments

No Known Activations