INDEX

Explanations

the word "as" followed by attributions or explanations

oai_token-act-pair · gpt-3.5-turbo

the phrase "as a" followed by various roles or identities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

redits

-0.82

rences

-0.81

eworks

-0.81

events

-0.80

DEV

-0.78

NetMessage

-0.77

pages

-0.77

CHAPTER

-0.76

files

-0.75

andon

-0.75

POSITIVE LOGITS

 result

1.53

 consequence

1.46

 reminder

1.28

 precaution

1.10

 refres

1.08

 matter

1.07

 consolation

1.04

 testament

1.00

 bonus

0.99

 lifelong

0.98

Activations Density 0.041%

the word "as" followed by attributions or explanations

the phrase "as a" followed by various roles or identities

No Comments

No Known Activations