INDEX

Explanations

the number "Two."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Two."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 catalog

-0.74

 reserves

-0.68

 greatly

-0.64

 actively

-0.63

 often

-0.62

 even

-0.62

ACT

-0.62

psy

-0.62

 liber

-0.61

 territory

-0.61

POSITIVE LOGITS

Two

3.10

Three

2.44

Four

2.31

Five

2.06

two

2.03

Eight

2.03

Two

2.02

Six

1.97

Seven

1.91

Nine

1.76

Activations Density 0.019%

the number "Two."

instances of the word "Two."

No Comments

No Known Activations