INDEX

Explanations

phrases related to negative character traits or actions, such as being a villain or a spy

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "in."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

Tickets

-0.74

 accordingly

-0.66

acters

-0.65

arios

-0.64

EH

-0.64

 includ

-0.63

rolet

-0.63

onso

-0.63

NOW

-0.62

POSITIVE LOGITS

 disguise

1.21

 sorts

0.84

 whose

0.82

 steroids

0.79

 nowhere

0.79

 miniature

0.78

 paradise

0.77

offensive

0.74

 spite

0.74

 nutshell

0.74

Activations Density 0.149%

phrases related to negative character traits or actions, such as being a villain or a spy

occurrences of the word "in."

No Comments

No Known Activations