INDEX

Explanations

the word "Paris" in various contexts

oai_token-act-pair · gpt-3.5-turbo

instances of a specific suffix or word pattern

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

 testimony

-0.68

 blast

-0.64

cl

-0.62

 dating

-0.61

 plastic

-0.59

 Radiation

-0.59

 Plastic

-0.59

 Dinosaur

-0.58

 compromising

-0.57

 firing

-0.57

POSITIVE LOGITS

ais

4.83

ois

1.26

ai

1.21

alis

1.13

aan

1.06

oir

1.02

aire

1.00

neau

1.00

aid

0.99

aque

0.98

Activations Density 0.012%

the word "Paris" in various contexts

instances of a specific suffix or word pattern

No Comments

No Known Activations