INDEX

Explanations

words related to physical locations, especially entrances and exits

oai_token-act-pair · gpt-3.5-turbo

references to entrances

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

amera

-0.72

enegger

-0.70

nian

-0.66

pend

-0.64

gans

-0.63

azeera

-0.63

die

-0.61

 hurd

-0.61

nutrition

-0.61

fortune

-0.61

POSITIVE LOGITS

 entrance

1.02

 entrances

0.97

 gates

0.89

prise

0.89

 thereto

0.88

pole

0.84

 doors

0.84

tained

0.83

tainment

0.80

stones

0.80

Activations Density 0.026%

words related to physical locations, especially entrances and exits

references to entrances

No Comments

No Known Activations