INDEX

Explanations

news-related terms

oai_token-act-pair · gpt-3.5-turbo

instances of the word "news" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Mississ

-0.72

warm

-0.71

ippery

-0.69

sei

-0.68

BuyableInstoreAndOnline

-0.68

nington

-0.67

Â¯Â¯

-0.66

ause

-0.65

volent

-0.65

aughs

-0.65

POSITIVE LOGITS

reader

0.98

room

0.91

worthy

0.90

feed

0.90

 aggregation

0.83

worthiness

0.82

agent

0.82

flash

0.82

Reporting

0.81

NEWS

0.79

Activations Density 0.035%

news-related terms

instances of the word "news" in various contexts

No Comments

No Known Activations