INDEX

Explanations

phrases related to physical objects being stacked or accumulated

oai_token-act-pair · gpt-3.5-turbo

references to physical collections or groups of objects

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

otiation

-0.72

estamp

-0.67

Interstitial

-0.66

oral

-0.66

 obser

-0.65

IAL

-0.65

hips

-0.64

nton

-0.64

clinton

-0.63

saf

-0.63

POSITIVE LOGITS

 piles

1.04

 piling

1.00

 pile

0.99

bags

0.91

 piled

0.88

wagon

0.74

bag

0.72

endium

0.72

bilt

0.70

 loads

0.70

Activations Density 0.020%

phrases related to physical objects being stacked or accumulated

references to physical collections or groups of objects

No Comments

No Known Activations