INDEX

Explanations

instances where something is being glimpsed or revealed briefly

oai_token-act-pair · gpt-3.5-turbo

instances of the word "peek" or variations of it

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

iction

-0.83

raid

-0.74

versive

-0.72

ordinate

-0.71

Interstitial

-0.70

dependent

-0.70

hate

-0.70

famous

-0.70

million

-0.69

relative

-0.69

POSITIVE LOGITS

 peek

1.45

 glimps

0.81

 wink

0.81

 hints

0.81

ileaks

0.78

 Nieto

0.76

udo

0.75

 glimpse

0.75

 headlights

0.74

 glances

0.71

Activations Density 0.003%

instances where something is being glimpsed or revealed briefly

instances of the word "peek" or variations of it

No Comments

No Known Activations