INDEX

Explanations

dates and descriptions related to news articles or events

oai_token-act-pair · gpt-3.5-turbo

dates and file references in a document

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ãĥ¼ãĥĨ

-0.70

 crossings

-0.67

 nearest

-0.65

Methods

-0.62

 loopholes

-0.61

killers

-0.61

 Traps

-0.60

=~

-0.60

cients

-0.60

========

-0.59

POSITIVE LOGITS

 file

1.16

 photo

1.12

 illustration

1.09

 photograph

1.05

 portrait

0.99

 FILE

0.98

file

0.98

FILE

0.96

File

0.92

 picture

0.90

Activations Density 0.069%

dates and descriptions related to news articles or events

dates and file references in a document

No Comments

No Known Activations