INDEX

Explanations

phrases containing the character '&' followed by other characters

oai_token-act-pair · gpt-3.5-turbo

conjunctions and phrases indicating connections between ideas or entities

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-0.81

bably

-0.77

etheless

-0.75

awaru

-0.73

thodox

-0.70

ivalent

-0.66

yip

-0.64

 Canaver

-0.62

theless

-0.61

 unusually

-0.59

POSITIVE LOGITS

3.27

2.08

(&

1.77

et

1.17

etc

1.16

Vs

1.16

(&

1.12

vs

1.12

1.10

AND

1.07

Activations Density 0.021%

phrases containing the character '&' followed by other characters

conjunctions and phrases indicating connections between ideas or entities

No Comments

No Known Activations