INDEX

Explanations

phrases or words related to potential danger or harm

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate various types of risk or danger

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

lins

-0.70

anche

-0.61

icut

-0.61

cest

-0.61

 Glock

-0.60

blem

-0.60

encers

-0.60

urd

-0.60

 awoken

-0.60

ricks

-0.60

POSITIVE LOGITS

EStream

0.86

groups

0.72

face

0.71

taker

0.70

taking

0.67

ulkan

0.67

 mosqu

0.66

 pestic

0.65

 harbour

0.64

________________________

0.64

Activations Density 0.010%

phrases or words related to potential danger or harm

phrases that indicate various types of risk or danger

No Comments

No Known Activations