INDEX

Explanations

phrases related to dangerous situations or negative events

oai_token-act-pair · gpt-3.5-turbo

mentions of cliffs

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

arers

-0.71

ammad

-0.69

Express

-0.69

inator

-0.68

IDs

-0.65

Abd

-0.64

 Organ

-0.64

arel

-0.63

tti

-0.63

 Organisation

-0.61

POSITIVE LOGITS

 cliff

4.03

 cliffs

2.57

 Cliff

1.92

 ledge

1.48

 canyon

1.39

 precip

1.35

 hill

1.32

 rocky

1.20

 waterfall

1.18

 abyss

1.16

Activations Density 0.013%

phrases related to dangerous situations or negative events

mentions of cliffs

No Comments

No Known Activations