INDEX

Explanations

references to government organizations or specific institutions, particularly related to health and medical services

oai_token-act-pair · gpt-3.5-turbo

mentions of health organizations and institutions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 favour

-0.76

peg

-0.68

essen

-0.67

 honour

-0.67

 genuine

-0.66

asar

-0.62

 upset

-0.62

 realised

-0.61

adow

-0.61

 disappointment

-0.61

POSITIVE LOGITS

 Centers

3.76

CDC

2.59

CDC

2.15

Cent

1.40

HHS

1.30

 Center

1.30

 Pediatrics

1.22

NIH

1.22

 Diagn

1.21

Center

1.21

Activations Density 0.015%

references to government organizations or specific institutions, particularly related to health and medical services

mentions of health organizations and institutions

No Comments

No Known Activations