INDEX

Explanations

words related to horrific and violent events

oai_token-act-pair · gpt-3.5-turbo

references to extreme violence or trauma

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

cius

-0.68

pai

-0.68

pta

-0.66

arten

-0.66

sama

-0.66

arton

-0.65

 Folder

-0.64

annis

-0.64

VO

-0.63

ritis

-0.63

POSITIVE LOGITS

ally

1.04

 earthqu

0.93

 atrocities

0.81

 horrors

0.81

 nightmares

0.76

 eleph

0.74

iously

0.73

 asylum

0.73

 beasts

0.72

 tort

0.72

Activations Density 0.027%

words related to horrific and violent events

references to extreme violence or trauma

No Comments

No Known Activations