INDEX

Explanations

phrases related to inaccuracies, errors, or unintended actions

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "accidental" events or awareness

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

PB

-0.77

geist

-0.76

senal

-0.70

SHIP

-0.69

 hope

-0.68

JV

-0.68

zhen

-0.67

SG

-0.66

ger

-0.63

hua

-0.62

POSITIVE LOGITS

redited

1.28

uracy

1.28

identally

1.06

urate

1.06

idental

1.04

reditation

1.01

acc

0.98

ustomed

0.96

acia

0.96

urrency

0.95

Activations Density 0.004%

phrases related to inaccuracies, errors, or unintended actions

references to the concept of "accidental" events or awareness

No Comments

No Known Activations