INDEX

Explanations

instances of the word "annual" followed by a number or specific context

oai_token-act-pair · gpt-3.5-turbo

references to annual reports or annual occurrences

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

wright

-0.89

enes

-0.84

enton

-0.73

uve

-0.70

arnaev

-0.69

hammad

-0.68

Ì¶

-0.68

holes

-0.68

cells

-0.68

arcity

-0.67

POSITIVE LOGITS

ized

1.26

ised

1.03

 dues

0.88

 turnover

0.84

ization

0.83

 pilgrimage

0.81

ize

0.81

 appropriations

0.80

 report

0.80

 occurrence

0.79

Activations Density 0.063%

instances of the word "annual" followed by a number or specific context

references to annual reports or annual occurrences

No Comments

No Known Activations