INDEX

Explanations

the word "can" with a high level of activation

oai_token-act-pair · gpt-3.5-turbo

occasions or expressions of effort and capability

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

furt

-0.65

UR

-0.63

 Lauder

-0.62

 rejection

-0.61

pend

-0.60

eur

-0.59

EUR

-0.58

 Politics

-0.58

IB

-0.58

prints

-0.56

POSITIVE LOGITS

 muster

1.08

 afford

1.08

berra

1.00

 feas

0.99

't

0.92

 convince

0.86

 possibly

0.86

adian

0.84

 safely

0.82

successfully

0.82

Activations Density 0.070%

the word "can" with a high level of activation

occasions or expressions of effort and capability

No Comments

No Known Activations