INDEX

Explanations

instances of the word "mis" followed by other letters, indicating the detection of words related to various forms of miscommunication, mismanagement, and mistake

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "mis-" indicating incorrectness or wrongdoing

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 outdoors

-0.64

 Rush

-0.64

ILA

-0.63

 opponents

-0.62

 extension

-0.61

 rush

-0.61

 extinct

-0.60

 expired

-0.59

 scraping

-0.58

 Skydragon

-0.58

POSITIVE LOGITS

cellaneous

1.50

beh

1.42

direction

1.42

appropri

1.39

informed

1.38

behavior

1.37

anthrop

1.36

directed

1.36

aligned

1.34

management

1.34

Activations Density 0.012%

instances of the word "mis" followed by other letters, indicating the detection of words related to various forms of miscommunication, mismanagement, and mistake

instances of the prefix "mis-" indicating incorrectness or wrongdoing

No Comments

No Known Activations