INDEX

Explanations

pronouns referring to an idea or entity previously mentioned in the text

oai_token-act-pair · gpt-3.5-turbo

references to the subject "it" and variations of the phrase construction

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Toast

-0.71

 Siri

-0.68

 Polk

-0.67

 Eleven

-0.66

WARE

-0.66

TED

-0.66

 Sadd

-0.65

 Rockies

-0.64

friends

-0.63

DAY

-0.62

POSITIVE LOGITS

unes

1.08

 consists

1.07

 involves

1.05

 arises

1.04

 strengthens

1.03

 enables

1.02

 imposes

1.01

 facilitates

1.00

 requires

1.00

 generates

0.99

Activations Density 0.287%

pronouns referring to an idea or entity previously mentioned in the text

references to the subject "it" and variations of the phrase construction

No Comments

No Known Activations