INDEX

Explanations

names of individuals in official or professional roles

oai_token-act-pair · gpt-3.5-turbo

instances of job titles and positions in professional contexts

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

bleacher

-0.74

"]=>

-0.73

ibe

-0.72

NetMessage

-0.72

anism

-0.72

igraph

-0.72

infeld

-0.68

luaj

-0.66

akespeare

-0.65

 Reloaded

-0.64

POSITIVE LOGITS

 deputy

1.12

 bureau

1.07

 representative

1.04

 director

1.03

 secretary

1.03

 commissioner

1.03

 attorney

1.02

 solicitor

0.97

 assistant

0.96

 investigator

0.96

Activations Density 0.693%

names of individuals in official or professional roles

instances of job titles and positions in professional contexts

No Comments

No Known Activations