INDEX

Explanations

references to physical actions or states of being

oai_token-act-pair · gpt-3.5-turbo

references to crime and legal consequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 assorted

-0.51

 hopefully

-0.49

asionally

-0.47

 doubtless

-0.47

 unforgettable

-0.47

 undoubtedly

-0.47

itton

-0.46

asso

-0.45

atari

-0.45

catentry

-0.44

POSITIVE LOGITS

 anymore

1.73

nor

1.53

 whatsoever

1.15

 anywhere

1.07

 anything

1.05

any

1.01

 slightest

1.01

 anybody

0.99

yet

0.90

nor

0.89

Activations Density 1.707%

references to physical actions or states of being

references to crime and legal consequences

No Comments

No Known Activations