INDEX

Explanations

adjectives preceded by 'ab-'

oai_token-act-pair · gpt-3.5-turbo

prefixes that convey a sense of supporting or enabling something

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Cups

-0.69

 offshore

-0.65

 Prosper

-0.63

CMS

-0.60

 attorneys

-0.59

 Appendix

-0.59

 Maker

-0.59

Hag

-0.59

 electricity

-0.58

 reclaimed

-0.58

POSITIVE LOGITS

omination

1.18

usable

1.12

rid

1.12

raham

1.10

rog

1.08

ject

1.04

ounding

1.02

err

0.97

lest

0.97

acus

0.97

Activations Density 0.015%

adjectives preceded by 'ab-'

prefixes that convey a sense of supporting or enabling something

No Comments

No Known Activations