INDEX

Explanations

terms related to profanity or profane language

oai_token-act-pair · gpt-3.5-turbo

references to professional contexts or occupations

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

CAR

-0.67

 Pinball

-0.67

 Amend

-0.65

Adds

-0.64

akin

-0.64

osite

-0.63

 Rabbit

-0.62

 Avalon

-0.62

 Coil

-0.62

actly

-0.62

POSITIVE LOGITS

 prof

4.20

prof

2.17

Prof

1.61

 Prof

1.55

 profiling

1.33

exp

1.14

 obsc

1.07

 vulgar

1.02

 corpor

0.98

 perf

0.98

Activations Density 0.026%

terms related to profanity or profane language

references to professional contexts or occupations

No Comments

No Known Activations