INDEX

Explanations

words related to things that are not helpful or effective

oai_token-act-pair · gpt-3.5-turbo

terms describing ineffectiveness or lack of value

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

uba

-0.76

TM

-0.68

Own

-0.67

 Highlights

-0.64

atis

-0.63

umn

-0.62

roma

-0.62

awi

-0.62

atown

-0.62

Ro

-0.62

POSITIVE LOGITS

 useless

3.40

 worthless

2.70

 meaningless

2.45

 pointless

2.40

 ineffective

2.37

 futile

2.21

 unus

1.94

 powerless

1.91

 helpless

1.79

seless

1.74

Activations Density 0.025%

words related to things that are not helpful or effective

terms describing ineffectiveness or lack of value

No Comments

No Known Activations