INDEX

Explanations

references to the phrase "GO"

oai_token-act-pair · gpt-3.5-turbo

references to a specific train system or service named "GO."

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

acters

-0.67

 Horus

-0.66

dimension

-0.65

 intervals

-0.63

stood

-0.62

Hour

-0.61

ussen

-0.61

cipline

-0.61

 Jude

-0.61

rika

-0.59

POSITIVE LOGITS

GO

1.29

GO

1.09

XL

0.86

JO

0.86

VER

0.84

ethe

0.81

VE

0.79

vernment

0.79

OGR

0.79

 unemploy

0.78

Activations Density 0.006%

references to the phrase "GO"

references to a specific train system or service named "GO."

No Comments

No Known Activations