INDEX

Explanations

names of people working in different roles and organizations

oai_token-act-pair · gpt-3.5-turbo

references to individuals' professional titles and affiliations

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

abiding

-0.57

 perfect

-0.56

awa

-0.55

 chilling

-0.53

 quicker

-0.50

 positives

-0.50

 tantal

-0.50

 horrend

-0.50

 tremend

-0.49

 crap

-0.49

POSITIVE LOGITS

ozo

0.61

 Petroleum

0.59

 Higher

0.59

omics

0.56

Dialogue

0.55

Bay

0.54

 Brands

0.54

Bio

0.54

Me

0.54

Brook

0.53

Activations Density 1.367%

names of people working in different roles and organizations

references to individuals' professional titles and affiliations

No Comments

No Known Activations