INDEX

Explanations

statements where someone is being validated or confirmed as correct

oai_token-act-pair · gpt-3.5-turbo

statements affirming correctness or agreement

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

earance

-0.65

 Gong

-0.62

cano

-0.62

irts

-0.62

battle

-0.61

gery

-0.60

nets

-0.60

 stability

-0.59

untu

-0.59

ains

-0.59

POSITIVE LOGITS

footed

0.94

eous

0.84

sighted

0.78

terday

0.72

Osw

0.72

 ignorant

0.68

 smack

0.68

fully

0.67

utherford

0.66

eyed

0.65

Activations Density 0.069%

statements where someone is being validated or confirmed as correct

statements affirming correctness or agreement

No Comments

No Known Activations