INDEX

Explanations

words related to emphasizing truth or reality

oai_token-act-pair · gpt-3.5-turbo

affirmations or confirmations of assertions being made

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 surely

-0.67

 likewise

-0.63

 newsletters

-0.62

FW

-0.62

imble

-0.61

Lastly

-0.60

olithic

-0.59

Finally

-0.58

HCR

-0.58

ÃĥÃĤ

-0.58

POSITIVE LOGITS

 existed

0.72

 quite

0.70

 harming

0.68

arial

0.68

 WORK

0.68

 happening

0.66

idad

0.66

 beneficial

0.64

ordered

0.64

 happened

0.64

Activations Density 0.076%

words related to emphasizing truth or reality

affirmations or confirmations of assertions being made

No Comments

No Known Activations