INDEX

Explanations

words related to things that exist or continue for an extended period

oai_token-act-pair · gpt-3.5-turbo

terms related to prolonged presence or effects

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ramid

-0.79

icient

-0.76

urat

-0.71

portion

-0.71

rees

-0.71

*/(

-0.69

reme

-0.68

OH

-0.67

opathic

-0.66

aughtered

-0.66

POSITIVE LOGITS

 linger

0.92

noon

0.90

 beans

0.76

butt

0.74

 lingering

0.71

 suspense

0.70

ipop

0.68

 lasting

0.67

 awhile

0.67

 seiz

0.66

Activations Density 0.004%

words related to things that exist or continue for an extended period

terms related to prolonged presence or effects

No Comments

No Known Activations