INDEX

Explanations

names related to replacements or duplicates

oai_token-act-pair · gpt-3.5-turbo

words related to deception or imitation

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

çĦ

-0.79

terday

-0.76

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

-0.76

ä½¿

-0.76

 DRAG

-0.70

crop

-0.70

è£ħ

-0.70

ãĥīãĥ©ãĤ´ãĥ³

-0.68

 fabrication

-0.68

VEL

-0.68

POSITIVE LOGITS

acements

1.07

ition

1.00

abulary

0.99

solete

0.93

iment

0.90

itive

0.89

arus

0.88

ension

0.88

ensions

0.86

emonium

0.86

Activations Density 0.036%

names related to replacements or duplicates

words related to deception or imitation

No Comments

No Known Activations