INDEX

Explanations

dates with news agency names

oai_token-act-pair · gpt-3.5-turbo

markers of citations or references to news agency content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Frozen

-0.68

 Stout

-0.67

 Exhibition

-0.67

Clo

-0.64

 Penny

-0.62

 Downs

-0.61

Deg

-0.61

 Dock

-0.61

 Squadron

-0.60

 Stall

-0.60

POSITIVE LOGITS

WASHINGTON

0.93

NA

0.86

ccording

0.82

BBC

0.80

externalActionCode

0.80

POL

0.77

Reuters

0.76

Aust

0.76

rawdownloadcloneembedreportprint

0.76

IAL

0.75

Activations Density 0.129%

dates with news agency names

markers of citations or references to news agency content

No Comments

No Known Activations