INDEX

Explanations

mentions of specific car features

oai_token-act-pair · gpt-3.5-turbo

specific terms related to high-end materials or finishes

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

NetMessage

-0.80

wav

-0.70

ovember

-0.68

 sqor

-0.67

↵Âł

-0.64

vert

-0.64

eteen

-0.64

 incent

-0.62

ocular

-0.61

gray

-0.60

POSITIVE LOGITS

's

0.84

 knows

0.66

iasis

0.65

Aki

0.64

ATTLE

0.63

stadt

0.63

 overload

0.63

 Rossi

0.63

 understands

0.62

isen

0.61

Activations Density 0.000%

mentions of specific car features

specific terms related to high-end materials or finishes

No Comments

No Known Activations