INDEX

Explanations

time-related phrases and expressions

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

icken

-0.48

oppable

-0.42

adle

-0.42

agger

-0.41

ensitive

-0.40

 feared

-0.40

anced

-0.40

ahl

-0.39

Ã¯ve

-0.38

glomer

-0.38

POSITIVE LOGITS

 week

1.01

 month

0.94

 fortnight

0.92

 semester

0.89

 weekend

0.87

 weeks

0.85

 WEEK

0.83

 year

0.81

decade

0.80

 nights

0.80

Activations Density 0.380%

time-related phrases and expressions

No Comments

No Known Activations