INDEX

Explanations

phrases indicating attempts to demonstrate or confirm something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "prove."

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

Nav

-0.64

Caf

-0.62

och

-0.61

 Partnership

-0.58

 Committee

-0.56

AU

-0.56

itching

-0.56

irection

-0.56

Ecc

-0.55

 executive

-0.55

POSITIVE LOGITS

 prove

3.53

 proves

2.12

 proving

2.03

 proved

1.89

 demonstrate

1.70

 proven

1.57

 dispro

1.56

 proof

1.44

proven

1.40

 convince

1.39

Activations Density 0.015%

phrases indicating attempts to demonstrate or confirm something

instances of the word "prove."

No Comments

No Known Activations