INDEX

Explanations

words related to subsequent events or actions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "subsequently" indicating a sequence of events or actions

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

 basics

-0.66

big

-0.65

 moms

-0.58

big

-0.58

 Anything

-0.57

za

-0.57

 thing

-0.55

 Tough

-0.54

mom

-0.53

 Enough

-0.53

POSITIVE LOGITS

 subsequently

3.16

 thereafter

2.06

 subsequent

2.06

sequently

1.99

 later

1.97

 consequently

1.91

 afterwards

1.75

 promptly

1.71

later

1.69

 eventually

1.60

Activations Density 0.009%

words related to subsequent events or actions

instances of the word "subsequently" indicating a sequence of events or actions

No Comments

No Known Activations