INDEX

Explanations

words related to emotional reactions or states such as "shocked", "heartened", "gratified", "humbled"

oai_token-act-pair · gpt-3.5-turbo

expressions of strong emotions and reactions

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

 occurrence

-0.76

 divergence

-0.73

 pecul

-0.73

 contribution

-0.71

 arcs

-0.69

 legitimacy

-0.68

 inacc

-0.66

 similarity

-0.66

 transitions

-0.66

 interchange

-0.66

POSITIVE LOGITS

raged

1.21

alysed

1.20

azed

1.19

rified

1.17

cerned

1.16

ivated

1.15

iliated

1.12

racted

1.12

informed

1.11

employed

1.10

Activations Density 0.283%

words related to emotional reactions or states such as "shocked", "heartened", "gratified", "humbled"

expressions of strong emotions and reactions

No Comments

No Known Activations