INDEX

Explanations

words with the prefix 'u-' that might indicate an action or state

oai_token-act-pair · gpt-3.5-turbo

occurrences of the letter 'u'

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 defect

-0.66

 executions

-0.62

 rendition

-0.61

 Citation

-0.61

 Strange

-0.59

 Alice

-0.59

 Bond

-0.56

 Sparrow

-0.56

 Attribution

-0.56

 proportional

-0.55

POSITIVE LOGITS

pperc

1.54

mpire

1.34

reth

1.29

ppers

1.18

seless

1.11

zbek

1.06

rolog

1.05

berman

1.04

mp

1.03

mph

1.02

Activations Density 0.013%

words with the prefix 'u-' that might indicate an action or state

occurrences of the letter 'u'

No Comments

No Known Activations