INDEX

Explanations

phrases related to completing a task or reaching a goal

oai_token-act-pair · gpt-3.5-turbo

instances of achievement or completion

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

ials

-0.64

 Slate

-0.63

 fluct

-0.62

Spread

-0.62

 constantly

-0.61

Rue

-0.60

conom

-0.58

flat

-0.58

 Scrib

-0.58

 alone

-0.56

POSITIVE LOGITS

 relent

0.75

 resil

0.74

atown

0.71

onso

0.70

iflower

0.69

aterasu

0.68

 reckoning

0.68

ppo

0.67

vana

0.66

ileaks

0.65

Activations Density 0.155%

phrases related to completing a task or reaching a goal

instances of achievement or completion

No Comments

No Known Activations