INDEX

Explanations

terms related to negative events or circumstances characterized by a decrease or reduction

oai_token-act-pair · gpt-3.5-turbo

references to "downs" in various contexts, particularly indicating declines or reductions

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

 statement

-0.63

 vision

-0.59

 alcoholic

-0.58

bat

-0.57

 account

-0.57

 Statement

-0.57

 possessed

-0.56

 shape

-0.56

 affidavit

-0.56

 article

-0.54

POSITIVE LOGITS

downs

4.48

down

2.29

 downs

1.88

 Downs

1.70

DOWN

1.67

Down

1.53

ups

1.41

offs

1.39

backs

1.32

overs

1.23

Activations Density 0.013%

terms related to negative events or circumstances characterized by a decrease or reduction

references to "downs" in various contexts, particularly indicating declines or reductions

No Comments

No Known Activations