INDEX

Explanations

names, in this case probably specifically the name "Victor" or "Victoria"

oai_token-act-pair · gpt-3.5-turbo

references to victims and victimization

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

clips

-0.76

 overhead

-0.72

 accompanying

-0.69

WAYS

-0.69

OPLE

-0.66

ó

-0.65

 interfering

-0.65

 tightening

-0.65

 habit

-0.64

 organic

-0.63

POSITIVE LOGITS

Vict

1.43

vict

1.34

orian

0.96

oire

0.96

ims

0.96

inel

0.86

 vict

0.85

ician

0.84

orically

0.82

ime

0.81

Activations Density 0.009%

names, in this case probably specifically the name "Victor" or "Victoria"

references to victims and victimization

No Comments

No Known Activations