INDEX

Explanations

the word "inmate"

oai_token-act-pair · gpt-3.5-turbo

references to inmates and incarceration

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

rous

-0.74

orie

-0.70

orically

-0.70

OPA

-0.70

drive

-0.69

Dri

-0.67

POL

-0.66

opic

-0.66

 Hague

-0.63

ku

-0.63

POSITIVE LOGITS

 inmates

1.06

 inmate

0.94

icts

0.91

 incarcerated

0.83

 closet

0.80

iaries

0.78

 sentenced

0.78

 detainees

0.75

 prisoners

0.74

 informants

0.72

Activations Density 0.016%

the word "inmate"

references to inmates and incarceration

No Comments

No Known Activations