INDEX

Explanations

news or media-related terms or content, such as "WATCH" and "NEWS."

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "WATCH" and related keywords indicating video content

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

agement

-0.78

assian

-0.74

fin

-0.71

agers

-0.70

Else

-0.70

aged

-0.69

ned

-0.68

roman

-0.67

oku

-0.66

ains

-0.63

POSITIVE LOGITS

WATCH

1.06

èĥ

1.05

ABLE

1.03

OWER

1.02

 VIDEOS

1.01

RESULTS

0.98

 IMAGES

0.96

LOCK

0.96

PORT

0.96

LEASE

0.95

Activations Density 0.028%

news or media-related terms or content, such as "WATCH" and "NEWS."

occurrences of the word "WATCH" and related keywords indicating video content

No Comments

No Known Activations