INDEX

Explanations

profane language, specifically the word "shit"

oai_token-act-pair · gpt-3.5-turbo

occurrences of strong negative expressions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

gary

-0.81

HCR

-0.80

================================

-0.74

PsyNetMessage

-0.73

fman

-0.70

AUT

-0.69

 Expend

-0.67

NetMessage

-0.67

BIL

-0.65

Minor

-0.63

POSITIVE LOGITS

shit

0.99

heads

0.93

storm

0.93

lord

0.92

bags

0.92

wit

0.90

loads

0.90

faced

0.90

lords

0.88

shots

0.86

Activations Density 0.014%

profane language, specifically the word "shit"

occurrences of strong negative expressions

No Comments

No Known Activations