INDEX

Explanations

phrases related to lists of items, often separated by "&"

oai_token-act-pair · gpt-3.5-turbo

references to different types of entities or categories often represented with an ampersand symbol

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

-0.79

bably

-0.77

awaru

-0.77

etheless

-0.70

thodox

-0.67

ivalent

-0.65

yip

-0.64

 Canaver

-0.61

TPPStreamerBot

-0.60

theless

-0.60

POSITIVE LOGITS

3.28

2.11

(&

1.84

(&

1.19

et

1.18

etc

1.18

Vs

1.13

vs

1.13

 thru

1.10

Activations Density 0.027%

phrases related to lists of items, often separated by "&"

references to different types of entities or categories often represented with an ampersand symbol

No Comments

No Known Activations