INDEX

Explanations

references to a year enclosed in parentheses

oai_token-act-pair · gpt-3.5-turbo

instances of parentheses or related punctuation

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

 lull

-0.75

 nuts

-0.70

ĸļ

-0.64

 psychiat

-0.63

 inund

-0.62

nesday

-0.62

 obscurity

-0.62

 everyday

-0.61

ylum

-0.61

 electr

-0.60

POSITIVE LOGITS

?)

1.06

formerly

1.06

including

1.01

...)

1.00

see

1.00

also

0.98

!)

0.98

excluding

0.97

via

0.96

emphasis

0.94

Activations Density 0.123%

references to a year enclosed in parentheses

instances of parentheses or related punctuation

No Known Activations