INDEX

Explanations

expressions where the speaker is making a noteworthy or emphasized statement

oai_token-act-pair · gpt-3.5-turbo

phrases that use the word "say" to express opinions or statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

fram

-0.62

Ku

-0.59

ipel

-0.57

onut

-0.57

abwe

-0.55

poon

-0.55

velength

-0.54

 timet

-0.54

edited

-0.53

TAM

-0.53

POSITIVE LOGITS

 goodbye

1.20

 hello

0.81

ysis

0.76

 aloud

0.74

 farewell

0.74

 definitively

0.71

 sorry

0.70

 Goodbye

0.68

hi

0.67

 thank

0.66

Activations Density 0.064%

expressions where the speaker is making a noteworthy or emphasized statement

phrases that use the word "say" to express opinions or statements

No Comments

No Known Activations