INDEX

Explanations

words related to control or being controlled

oai_token-act-pair · gpt-3.5-turbo

references to control or regulation

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

stad

-0.80

reads

-0.79

offer

-0.73

ennes

-0.68

wl

-0.68

jen

-0.67

fare

-0.66

baugh

-0.65

anth

-0.65

bishop

-0.62

POSITIVE LOGITS

 controlled

3.88

 Controlled

2.74

controlled

2.73

 controlling

1.96

 uncontrolled

1.83

 control

1.77

 contro

1.76

 controls

1.55

control

1.52

 governed

1.51

Activations Density 0.026%

words related to control or being controlled

references to control or regulation

No Comments

No Known Activations