INDEX

Explanations

the word "mini" followed by a number indicating a scale or version

oai_token-act-pair · gpt-3.5-turbo

references to anything described as "mini."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

âķĲâķĲ

-0.81

Ö¼

-0.76

yer

-0.72

rer

-0.68

lain

-0.68

 Bleach

-0.66

REAM

-0.65

ILLE

-0.63

 Grind

-0.62

 depl

-0.62

POSITIVE LOGITS

ature

1.19

atur

1.18

atures

0.99

ocular

0.93

atomic

0.80

 sized

0.77

eco

0.75

ota

0.75

oby

0.73

mini

0.72

Activations Density 0.009%

the word "mini" followed by a number indicating a scale or version

references to anything described as "mini."

No Comments

No Known Activations