INDEX

Explanations

words related to extremely negative or horrifying events

oai_token-act-pair · gpt-3.5-turbo

references to extreme negative events or situations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

annis

-0.79

pai

-0.73

sers

-0.73

arten

-0.72

verning

-0.71

pta

-0.71

utenant

-0.69

sama

-0.69

cius

-0.68

glas

-0.68

POSITIVE LOGITS

 horrors

0.86

 asylum

0.84

 atrocities

0.83

ally

0.82

 ordeal

0.80

 earthqu

0.77

omic

0.76

 tales

0.75

 nightmares

0.75

 imagery

0.74

Activations Density 0.053%

words related to extremely negative or horrifying events

references to extreme negative events or situations

No Comments

No Known Activations