INDEX

Explanations

words related to governmental and political entities

oai_token-act-pair · gpt-3.5-turbo

sequences that indicate entities or organizations within a document

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

antz

-0.66

 conversions

-0.64

ocalypse

-0.64

 mats

-0.63

 wrinkles

-0.62

 reversible

-0.61

stract

-0.60

aukee

-0.60

 heights

-0.60

 demos

-0.59

POSITIVE LOGITS

 meanwhile

1.31

 headquartered

1.26

 which

1.16

 chaired

1.09

 whose

1.07

 founded

1.06

 citing

1.04

which

1.03

whose

0.97

 however

0.94

Activations Density 0.153%

words related to governmental and political entities

sequences that indicate entities or organizations within a document

No Comments

No Known Activations