INDEX

Explanations

the word "pleaded."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "plead" and its variations related to legal contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

lav

-0.76

ONES

-0.76

grad

-0.69

 Bohem

-0.66

java

-0.65

rats

-0.63

aged

-0.63

 Chop

-0.63

edition

-0.63

odor

-0.63

POSITIVE LOGITS

 plead

1.37

 pleaded

1.20

 pleading

1.09

 pled

1.08

 plea

0.92

 pleas

0.88

MENTS

0.82

 guilty

0.77

geoning

0.76

 barg

0.74

Activations Density 0.008%

the word "pleaded."

instances of the word "plead" and its variations related to legal contexts

No Comments

No Known Activations