INDEX

Explanations

phrases related to news and updates

oai_token-act-pair · gpt-3.5-turbo

positive news or good updates

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

furt

-0.75

asus

-0.69

amera

-0.68

arget

-0.68

avorable

-0.67

pload

-0.66

VI

-0.66

JP

-0.65

inton

-0.65

PRES

-0.64

POSITIVE LOGITS

 afterlife

0.69

 outweigh

0.69

 angels

0.67

iest

0.66

 Rubin

0.65

 outdoors

0.63

 classics

0.63

 underside

0.63

 aisle

0.63

 goodness

0.63

Activations Density 0.409%

phrases related to news and updates

positive news or good updates

No Comments

No Known Activations