INDEX

Explanations

instances where something is being replaced with something else

oai_token-act-pair · gpt-3.5-turbo

instances of the word "replaced"

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ively

-0.73

 firsthand

-0.63

 Scripture

-0.58

 spectrum

-0.58

bies

-0.57

ities

-0.56

imeter

-0.56

 diligence

-0.56

jud

-0.56

 Thursday

-0.56

POSITIVE LOGITS

 replaced

3.87

 replaces

2.37

 substituted

2.18

 replace

2.17

 replacing

2.10

 suppl

1.74

 replacement

1.73

 Replace

1.70

 replacements

1.69

 swapped

1.68

Activations Density 0.014%

instances where something is being replaced with something else

instances of the word "replaced"

No Comments

No Known Activations