INDEX

Explanations

security and defense-related terms

oai_token-act-pair · gpt-3.5-turbo

references to groups of people, particularly in the context of specific roles or professions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

osterone

-0.73

ebook

-0.66

umerable

-0.66

Market

-0.65

theless

-0.64

inventoryQuantity

-0.64

à¨

-0.62

à¤

-0.59

optim

-0.59

orically

-0.59

POSITIVE LOGITS

who

1.20

 whom

1.18

 interviewed

1.06

 indicted

1.06

hips

1.04

 vying

0.98

 participating

0.97

 apiece

0.96

 whose

0.95

 involved

0.94

Activations Density 0.253%

security and defense-related terms

references to groups of people, particularly in the context of specific roles or professions

No Comments

No Known Activations