INDEX

Explanations

scientific journal names

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

references to the journal 'Nature' and its associated publications

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

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

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_post

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

haul

-0.66

 perjury

-0.63

%-

-0.63

tru

-0.60

 thee

-0.60

strap

-0.60

 stale

-0.60

 hump

-0.59

elf

-0.58

%);

-0.58

POSITIVE LOGITS

 Communications

1.17

 Genetics

1.01

 Methods

0.99

 Letters

0.98

Adv

0.96

 Communication

0.95

 Reviews

0.93

 Neuroscience

0.92

 Materials

0.92

 Chemistry

0.92

Activations Density 0.031%

scientific journal names

references to the journal 'Nature' and its associated publications

No Comments

No Known Activations