INDEX

Explanations

written accounts related to criminal activities, emergencies, and law enforcement

oai_token-act-pair · gpt-3.5-turbo

references to police and crime-related activities

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

Craft

-0.73

ptions

-0.73

ãĤ»

-0.71

Trade

-0.63

EEK

-0.62

Percent

-0.61

antasy

-0.60

imum

-0.60

Economic

-0.60

POSITIVE LOGITS

 brig

0.84

 distraught

0.83

 barric

0.79

 altercation

0.75

 NYPD

0.73

 unmarked

0.73

 sheriff

0.73

 saddened

0.73

 downed

0.72

 gunman

0.72

Activations Density 3.117%

written accounts related to criminal activities, emergencies, and law enforcement

references to police and crime-related activities

No Comments

No Known Activations