INDEX

Explanations

mentions of layers, especially those described in terms of depth or intensity

oai_token-act-pair · gpt-3.5-turbo

references to layers, particularly in a context of construction or complexity

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

orie

-0.81

iggurat

-0.77

etheus

-0.73

date

-0.72

itarian

-0.71

ãĥīãĥ©

-0.69

oris

-0.69

ivia

-0.68

 Springer

-0.68

ILLE

-0.67

POSITIVE LOGITS

 thickness

1.26

 layers

1.23

 layer

1.08

layer

0.85

 thicker

0.79

 underneath

0.79

 coats

0.78

arity

0.78

coat

0.78

 veil

0.77

Activations Density 0.043%

mentions of layers, especially those described in terms of depth or intensity

references to layers, particularly in a context of construction or complexity

No Comments

No Known Activations