INDEX

Explanations

words related to something being replaced, succeeded, or outperformed by someone or something else

oai_token-act-pair · gpt-3.5-turbo

references to "predecessors" or previous entities in any context

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

oros

-0.80

arov

-0.79

heart

-0.70

agra

-0.69

rage

-0.68

alle

-0.68

erer

-0.66

ux

-0.66

Alien

-0.66

osher

-0.66

POSITIVE LOGITS

 predecessor

1.03

 predecessors

1.03

cules

0.82

Tad

0.80

 successor

0.74

 generations

0.73

 incarnation

0.71

 incarn

0.69

thood

0.69

 successors

0.67

Activations Density 0.012%

words related to something being replaced, succeeded, or outperformed by someone or something else

references to "predecessors" or previous entities in any context

No Comments

No Known Activations