INDEX

Explanations

words related to evaluation and rating

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate ratings or evaluations of entities

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

arch

-0.65

 stool

-0.64

pora

-0.63

 confusion

-0.63

arte

-0.62

anse

-0.62

aye

-0.61

Alm

-0.60

 scattering

-0.60

minist

-0.60

POSITIVE LOGITS

 rated

4.09

 Rated

2.55

 rating

1.95

rated

1.94

 ranked

1.91

Rated

1.85

 graded

1.71

 ratings

1.70

 Ranked

1.48

Rating

1.46

Activations Density 0.002%

words related to evaluation and rating

phrases that indicate ratings or evaluations of entities

No Comments

No Known Activations