INDEX

Explanations

phrases starting with "We" that suggest statements, decisions, or actions

oai_token-act-pair · gpt-3.5-turbo

repeated phrases emphasizing collective statements or beliefs

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

Ore

-0.64

mund

-0.63

 Eleven

-0.62

 Haku

-0.61

 Wikipedia

-0.60

Appearances

-0.60

Yor

-0.57

 Harding

-0.57

Posts

-0.57

reddits

-0.56

POSITIVE LOGITS

're

1.41

've

1.36

akening

1.20

'll

1.14

eks

1.13

ighed

1.11

 believe

1.08

 ourselves

1.07

 intend

1.05

athered

1.05

Activations Density 0.196%

phrases starting with "We" that suggest statements, decisions, or actions

repeated phrases emphasizing collective statements or beliefs

No Comments

No Known Activations