INDEX

Explanations

the word "extra."

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "extra" or supplementary items

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

theless

-0.85

\\\\\\\\

-0.74

better

-0.69

 CrossRef

-0.66

Hes

-0.66

-0.64

 Cosponsors

-0.64

geries

-0.63

iage

-0.63

eer

-0.62

POSITIVE LOGITS

judicial

1.39

ordinary

1.36

verted

1.06

veh

1.01

ord

0.93

 virgin

0.87

ngth

0.86

verts

0.86

vas

0.85

 goodies

0.79

Activations Density 0.034%

the word "extra."

references to the concept of "extra" or supplementary items

No Comments

No Known Activations