INDEX

Explanations

terms related to specific names or keywords, such as "Potomac" or "Potter"

oai_token-act-pair · gpt-3.5-turbo

references to the word "Pot" and its variations in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

anwhile

-0.79

 preached

-0.66

 detached

-0.65

dress

-0.64

âĸ¬

-0.63

Ń·

-0.63

draft

-0.62

 Witness

-0.62

 Defenders

-0.60

 silence

-0.60

POSITIVE LOGITS

atoes

1.50

assium

1.40

entially

1.09

hesda

1.01

Pot

0.99

luck

0.97

entials

0.97

onga

0.94

Pot

0.89

ato

0.89

Activations Density 0.007%

terms related to specific names or keywords, such as "Potomac" or "Potter"

references to the word "Pot" and its variations in various contexts

No Comments

No Known Activations