INDEX

Explanations

phrases related to question and answer formats

oai_token-act-pair · gpt-3.5-turbo

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ufact

-0.75

raint

-0.68

etheless

-0.60

fell

-0.57

ipedia

-0.57

wolf

-0.56

 Goldman

-0.56

fitted

-0.56

fitting

-0.54

neb

-0.53

POSITIVE LOGITS

FAQ

0.91

Answer

0.90

0.88

Reply

0.85

Ds

0.79

Cs

0.78

answered

0.78

answer

0.78

 Answer

0.76

LD

0.75

Activations Density 0.022%

phrases related to question and answer formats

No Comments

No Known Activations