INDEX

Explanations

references to directions, colors, days of the week, and sizes

oai_token-act-pair · gpt-3.5-turbo

words and concepts related to directional and positional references

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

mun

-0.74

-0.70

anan

-0.67

arb

-0.66

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

-0.63

 Machines

-0.63

 RELE

-0.63

appropri

-0.60

kef

-0.60

ando

-0.60

POSITIVE LOGITS

itto

0.78

 counterpart

0.74

 equivalents

0.72

 flakes

0.68

 Bulldogs

0.66

 Malays

0.64

tails

0.64

 counterparts

0.64

 coasts

0.63

 versa

0.63

Activations Density 0.800%

references to directions, colors, days of the week, and sizes

words and concepts related to directional and positional references

No Comments

No Known Activations