INDEX

Explanations

passive voice constructions

oai_token-act-pair · gpt-3.5-turbo

verbs associated with reporting, revealing, and estimating events or information

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

isode

-0.76

avorite

-0.64

vette

-0.62

ROM

-0.62

orio

-0.61

Exit

-0.61

ypes

-0.61

venant

-0.60

angs

-0.60

aping

-0.59

POSITIVE LOGITS

 unanimously

0.75

 raining

0.68

 that

0.64

 premature

0.62

 doubtful

0.61

 advisable

0.59

sonian

0.56

 downhill

0.56

 incidentally

0.55

 folly

0.55

Activations Density 0.092%

passive voice constructions

verbs associated with reporting, revealing, and estimating events or information

No Comments

No Known Activations