INDEX

Explanations

names and titles of specific individuals

oai_token-act-pair · gpt-3.5-turbo

references to individuals, particularly notable figures and their relationships

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 FIRE

-0.60

 ACTION

-0.58

risome

-0.56

 study

-0.56

 seism

-0.55

ogical

-0.51

 observational

-0.50

 seismic

-0.50

 rigorous

-0.50

ursday

-0.50

POSITIVE LOGITS

 Kardash

0.74

 whom

0.71

)."

0.71

).[

0.70

)</

0.67

 respectively

0.66

essor

0.66

çĶ°

0.66

etc

0.64

 heir

0.64

Activations Density 1.120%

names and titles of specific individuals

references to individuals, particularly notable figures and their relationships

No Comments

No Known Activations