INDEX

Explanations

causal relationships or implications between different elements of a narrative or discourse

oai_token-act-pair · gpt-3.5-turbo

instances of punctuation, specifically commas used in complex or lengthy sentences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ety

-0.66

¬¼

-0.64

UF

-0.60

enary

-0.54

herent

-0.54

§

-0.53

pec

-0.53

iple

-0.52

orn

-0.50

reth

-0.50

POSITIVE LOGITS

 albeit

1.08

 although

0.93

 namely

0.89

 whereas

0.85

 however

0.84

 though

0.84

 regardless

0.81

 which

0.79

but

0.79

huh

0.79

Activations Density 1.462%

causal relationships or implications between different elements of a narrative or discourse

instances of punctuation, specifically commas used in complex or lengthy sentences

No Comments

No Known Activations