INDEX

Explanations

phrases related to things being revealed or uncovered

oai_token-act-pair · gpt-3.5-turbo

instances of the word "unse" and its variations, as well as references to the name "Courtney."

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

iang

-0.76

iants

-0.72

managed

-0.67

 rigging

-0.65

%%%%

-0.65

schild

-0.63

 Cosponsors

-0.62

 Mongolia

-0.62

sson

-0.61

dial

-0.61

POSITIVE LOGITS

nces

0.92

ney

0.90

neys

0.87

nant

0.82

xy

0.80

tis

0.78

vy

0.77

î

0.75

ãĥ´

0.74

Filename

0.73

Activations Density 0.028%

phrases related to things being revealed or uncovered

instances of the word "unse" and its variations, as well as references to the name "Courtney."

No Comments

No Known Activations