INDEX

Explanations

instances of the word "watching."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "watching."

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

esan

-0.77

 misunderstanding

-0.77

phi

-0.77

cision

-0.73

orrect

-0.67

usable

-0.66

interstitial

-0.66

apolog

-0.66

enture

-0.64

oxy

-0.64

POSITIVE LOGITS

 Watching

1.30

 watching

1.17

 watched

1.04

wat

0.90

ÃįÃį

0.89

 watch

0.88

 watches

0.85

tower

0.84

 Watch

0.83

watching

0.83

Activations Density 0.016%

instances of the word "watching."

instances of the word "watching."

No Comments

No Known Activations