INDEX

Explanations

words related to medical conditions, particularly serious illnesses

oai_token-act-pair · gpt-3.5-turbo

references to illness and medical conditions

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

 compr

-0.78

pees

-0.75

 bloc

-0.71

ogle

-0.68

ramid

-0.67

zees

-0.66

 bumper

-0.65

reme

-0.61

de

-0.61

cycl

-0.61

POSITIVE LOGITS

 illness

0.89

 outbreaks

0.89

 outbreak

0.86

 illnesses

0.84

ousing

0.82

 plag

0.80

ochond

0.78

 onset

0.77

 symptoms

0.76

 Symptoms

0.75

Activations Density 0.014%

words related to medical conditions, particularly serious illnesses

references to illness and medical conditions

No Comments

No Known Activations