INDEX

Explanations

words ending in "der" with an emphasis on those ending in "der" followed by a high-activation digit

oai_token-act-pair · gpt-3.5-turbo

words related to individuals or people

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 African

-0.61

INTER

-0.59

 Echoes

-0.57

 elapsed

-0.56

 Korean

-0.56

zzo

-0.55

 Asian

-0.55

Pros

-0.55

 Impossible

-0.54

[|

-0.54

POSITIVE LOGITS

der

1.21

theless

1.12

cair

0.98

iving

0.97

rors

0.92

iver

0.91

igans

0.90

ricks

0.88

rine

0.87

mil

0.87

Activations Density 0.004%

words ending in "der" with an emphasis on those ending in "der" followed by a high-activation digit

words related to individuals or people

No Comments

No Known Activations