INDEX

Explanations

mentions of the word "Con" followed by a high number

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "Con" suggesting it looks for words related to connections or conversations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

OHN

-0.66

TPS

-0.65

wip

-0.64

 Rover

-0.64

Wem

-0.63

 practicable

-0.62

terday

-0.61

 hydra

-0.60

aceutical

-0.60

 Uzbek

-0.58

POSITIVE LOGITS

stant

1.26

ventions

1.15

cept

1.14

ference

1.12

secut

1.08

clusions

1.08

verting

1.06

cerning

1.04

verted

1.04

joined

1.04

Activations Density 0.014%

mentions of the word "Con" followed by a high number

instances of the prefix "Con" suggesting it looks for words related to connections or conversations

No Comments

No Known Activations