INDEX

Explanations

explicit language or profanity

oai_token-act-pair · gpt-3.5-turbo

instances of the word "shit" 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 @ 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

HCR

-0.74

fman

-0.71

 Expend

-0.71

arb

-0.67

 Somers

-0.66

PsyNetMessage

-0.65

ervation

-0.58

NetMessage

-0.57

Austral

-0.56

hani

-0.56

POSITIVE LOGITS

bags

1.02

heads

0.98

pants

0.93

loads

0.92

 detector

0.92

posts

0.92

storm

0.91

shots

0.90

lords

0.87

shit

0.85

Activations Density 0.040%

explicit language or profanity

instances of the word "shit" in various contexts

No Comments

No Known Activations