INDEX

Explanations

references to late-night television show titles, especially "Tonight Show"

oai_token-act-pair · gpt-3.5-turbo

mentions of the word "Tonight."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 annex

-0.65

mot

-0.65

utz

-0.64

 instruct

-0.63

 applic

-0.62

sub

-0.60

 realise

-0.60

 scope

-0.59

 encl

-0.58

åĤ

-0.58

POSITIVE LOGITS

 Tonight

3.20

 Today

0.98

 TODAY

0.92

Tonight

0.91

 Weekly

0.87

rums

0.82

 Reporting

0.74

 tonight

0.74

 Recorded

0.74

 Lowell

0.73

Activations Density 0.003%

references to late-night television show titles, especially "Tonight Show"

mentions of the word "Tonight."

No Comments

No Known Activations