INDEX

Explanations

words that include the prefix "ex-" along with references to relationships or previous connections

oai_token-act-pair · gpt-3.5-turbo

mentions of past relationships or individuals referred to as 'ex-' in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 imagination

-0.82

 shading

-0.81

 amplification

-0.80

 shroud

-0.76

 imitation

-0.75

 Illum

-0.75

 starve

-0.75

 dances

-0.75

 contextual

-0.74

 weighted

-0.73

POSITIVE LOGITS

girlfriend

1.81

employ

1.74

wife

1.66

husband

1.64

military

1.50

president

1.48

boss

1.47

wives

1.47

football

1.47

boy

1.45

Activations Density 0.027%

words that include the prefix "ex-" along with references to relationships or previous connections

mentions of past relationships or individuals referred to as 'ex-' in various contexts

No Comments

No Known Activations