INDEX

Explanations

words related to things that have come to an end or have been concluded

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "ended."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Gul

-0.61

 Phen

-0.60

 angel

-0.59

 ident

-0.58

 Gael

-0.58

Tex

-0.58

 disciple

-0.58

cyn

-0.56

 figure

-0.56

 mole

-0.56

POSITIVE LOGITS

ended

4.36

ends

2.69

ending

2.58

ENDED

2.43

end

2.05

enders

1.82

END

1.80

ender

1.64

ension

1.60

ensions

1.42

Activations Density 0.011%

words related to things that have come to an end or have been concluded

occurrences of the word "ended."

No Comments

No Known Activations