INDEX

Explanations

phrases related to exposing or revealing information

oai_token-act-pair · gpt-3.5-turbo

instances of the verb "is" as a key indicator of statements or assertions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

•Layer 1: Is/Was tokens

Fires on the token is

No Comments

Negative Logits

uese

-0.62

peg

-0.60

inav

-0.55

aturdays

-0.53

 slips

-0.52

 encounters

-0.51

 hatch

-0.50

 muster

-0.49

 Towers

-0.48

enture

-0.48

POSITIVE LOGITS

rael

0.96

nt

0.87

cussion

0.80

 hereby

0.76

senal

0.74

olate

0.74

 definitely

0.72

Ĥİ

0.71

 unlikely

0.70

claimer

0.69

Activations Density 0.396%

phrases related to exposing or revealing information

instances of the verb "is" as a key indicator of statements or assertions

No Comments

No Known Activations