INDEX

Explanations

television channel names followed by numerical values or specific program titles

oai_token-act-pair · gpt-3.5-turbo

references to news channels and their broadcasts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

 Â¥

-0.74

wikipedia

-0.73

estamp

-0.71

 seiz

-0.67

 flush

-0.67

 term

-0.66

netflix

-0.63

amsung

-0.61

apache

-0.61

 Alibaba

-0.61

POSITIVE LOGITS

 Productions

0.99

ONSORED

0.86

TV

0.84

 Networks

0.81

CTV

0.81

 Investig

0.80

 Broadcasting

0.78

 Television

0.76

 Studios

0.75

 crews

0.74

Activations Density 0.086%

television channel names followed by numerical values or specific program titles

references to news channels and their broadcasts

No Comments

No Known Activations