INDEX

Explanations

phrases discussing reflection or representation of something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "reflect" indicating a commentary on beliefs or values

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

amia

-0.79

ciating

-0.73

ccess

-0.72

cham

-0.67

arter

-0.67

FFFF

-0.66

cel

-0.64

FT

-0.63

FS

-0.62

duct

-0.61

POSITIVE LOGITS

 reflect

1.24

 reflections

1.11

 reflecting

1.10

 reflects

1.09

reflect

1.05

 reflection

1.00

 reflected

0.99

 Reflect

0.95

soType

0.90

 reflective

0.80

Activations Density 0.008%

phrases discussing reflection or representation of something

instances of the word "reflect" indicating a commentary on beliefs or values

No Comments

No Known Activations