INDEX

Explanations

instances of the word "watched" at varying activations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "watched."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

interstitial

-0.84

cision

-0.81

activation

-0.80

esan

-0.75

enture

-0.74

orrect

-0.74

 misunderstanding

-0.73

pex

-0.71

acies

-0.70

por

-0.68

POSITIVE LOGITS

 Watching

1.18

 watched

1.17

 watching

1.02

 watch

0.90

 watches

0.90

tower

0.88

 WATCH

0.84

wat

0.82

 aloud

0.81

=#

0.80

Activations Density 0.009%

instances of the word "watched" at varying activations

instances of the word "watched."

No Comments

No Known Activations