INDEX

Explanations

phrases related to interaction and communication between people, especially expressing opinions or reactions

oai_token-act-pair · gpt-3.5-turbo

mentions of public outrage or social media reactions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

senal

-0.78

 presided

-0.74

thood

-0.71

 overseen

-0.67

izontal

-0.66

è¦ļéĨĴ

-0.65

ainment

-0.63

acl

-0.61

partial

-0.61

Complete

-0.60

POSITIVE LOGITS

 clam

0.92

 themselves

0.86

 louder

0.85

 boycot

0.75

 vitri

0.72

 their

0.71

 overwhelmingly

0.70

their

0.69

 boycott

0.67

 uproar

0.67

Activations Density 0.446%

phrases related to interaction and communication between people, especially expressing opinions or reactions

mentions of public outrage or social media reactions

No Comments

No Known Activations