INDEX

Explanations

terms related to investigative work or journalism

oai_token-act-pair · gpt-3.5-turbo

references to investigative journalism or related investigative processes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĤ¼ãĤ¦ãĤ¹

-0.90

ults

-0.68

 yeast

-0.67

Origin

-0.65

ware

-0.65

lands

-0.63

 cups

-0.63

marriage

-0.62

XL

-0.62

 Roses

-0.61

POSITIVE LOGITS

 investigative

3.36

 Investigative

2.58

 investig

1.55

 undercover

1.30

 investigator

1.22

 journalistic

1.17

 Investig

1.12

 forensic

1.11

 Forensic

1.10

 Pulitzer

1.07

Activations Density 0.024%

terms related to investigative work or journalism

references to investigative journalism or related investigative processes

No Comments

No Known Activations