INDEX

Explanations

news-related terms such as names, social media references, and events

oai_token-act-pair · gpt-3.5-turbo

occurrences of numeric or counting terms in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

starting

-0.60

milo

-0.60

iations

-0.59

iod

-0.59

 branching

-0.58

 plaster

-0.58

 bould

-0.56

 resin

-0.55

 bombed

-0.55

eday

-0.55

POSITIVE LOGITS

 SHARES

1.09

 Shares

0.73

 Expand

0.70

Posted

0.69

PROV

0.67

CLUS

0.63

 Reporter

0.63

--+

0.63

 Tyson

0.63

 CITY

0.62

Activations Density 0.036%

news-related terms such as names, social media references, and events

occurrences of numeric or counting terms in the text

No Comments

No Known Activations