INDEX

Explanations

references to downgrading, depreciation, or reduction

oai_token-act-pair · gpt-3.5-turbo

instances of the word "down" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

andum

-0.68

-0.66

ãĥŁ

-0.56

Ã«

-0.56

 avail

-0.56

Discover

-0.55

¶ħ

-0.55

tein

-0.55

-0.54

 Pros

-0.54

POSITIVE LOGITS

graded

1.43

grading

1.30

LOAD

1.29

stairs

1.28

pour

1.14

grades

1.14

hill

1.06

loaded

1.03

grade

0.97

sized

0.96

Activations Density 0.045%

references to downgrading, depreciation, or reduction

instances of the word "down" in various contexts

No Comments

No Known Activations