INDEX

Explanations

mentions of completeness or fullness

oai_token-act-pair · gpt-3.5-turbo

instances of the word "full."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

Wars

-0.78

 Disciple

-0.78

Pac

-0.72

achu

-0.67

aughs

-0.66

Tup

-0.66

åĤ

-0.65

SHIP

-0.65

estern

-0.62

 Guys

-0.62

POSITIVE LOGITS

 fled

0.91

heartedly

0.87

ifully

0.84

 extent

0.84

 full

0.84

 brunt

0.82

erton

0.78

 fuller

0.74

width

0.72

inen

0.71

Activations Density 0.028%

mentions of completeness or fullness

instances of the word "full."

No Comments

No Known Activations