INDEX

Explanations

phrases related to avoiding or preventing negative outcomes

oai_token-act-pair · gpt-3.5-turbo

verbal forms related to overcoming challenges or difficulties

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

cdn

-0.65

 constitu

-0.63

 narrated

-0.63

 UNIVERS

-0.58

lag

-0.58

Exper

-0.57

 subsequ

-0.57

 bounded

-0.55

 craving

-0.55

ancies

-0.54

POSITIVE LOGITS

coat

0.90

·

0.84

¸

0.84

otom

0.82

otomy

0.81

ow

0.80

stock

0.80

Ń·

0.78

©¶æ

0.77

ħ

0.76

Activations Density 0.176%

phrases related to avoiding or preventing negative outcomes

verbal forms related to overcoming challenges or difficulties

No Comments

No Known Activations