INDEX

Explanations

keywords related to investigative reports and legal actions

oai_token-act-pair · gpt-3.5-turbo

phrases related to political and legal contexts

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

amorph

-0.72

viation

-0.71

etheless

-0.67

 FranÃ§

-0.66

theless

-0.63

 largeDownload

-0.63

 appell

-0.63

 "$:/

-0.63

techn

-0.63

vier

-0.62

POSITIVE LOGITS

 Kavanaugh

1.04

 Dems

0.75

 Rosenstein

0.70

Ħ¢

0.70

%]

0.70

 probe

0.66

 holiest

0.63

RTX

0.63

 Replay

0.61

Ford

0.61

Activations Density 0.062%

keywords related to investigative reports and legal actions

phrases related to political and legal contexts

No Comments

No Known Activations