INDEX

Explanations

the word "Returns."

oai_token-act-pair · gpt-3.5-turbo

references to movies or sequels

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

ãĤ¨ãĥ«

-0.78

 Vehicles

-0.78

ãĥīãĥ©

-0.74

Accessory

-0.74

CLASSIFIED

-0.73

 Warfare

-0.72

ITED

-0.70

WARE

-0.69

 reluct

-0.68

 AUTH

-0.65

POSITIVE LOGITS

 neoc

0.67

 candidacy

0.64

 that

0.63

ever

0.60

morrow

0.59

that

0.59

aker

0.58

 alphabet

0.58

interest

0.58

 olig

0.58

Activations Density 0.000%

the word "Returns."

references to movies or sequels

No Comments

No Known Activations