INDEX

Explanations

words related to potential conflicts or crises that are escalating

oai_token-act-pair · gpt-3.5-turbo

conditional statements about potential negative outcomes or escalations in conflicts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 researched

-0.69

 guessing

-0.68

 inspecting

-0.67

TAG

-0.66

 Registered

-0.66

athlon

-0.66

 researching

-0.65

 guesses

-0.65

Favorite

-0.65

 scanned

-0.65

POSITIVE LOGITS

 destabil

1.27

 entail

1.27

 weaken

1.27

 embold

1.26

 plunge

1.25

 undermine

1.24

 provoke

1.24

 culmin

1.23

 escalate

1.22

 exacerbate

1.21

Activations Density 0.211%

words related to potential conflicts or crises that are escalating

conditional statements about potential negative outcomes or escalations in conflicts

No Comments

No Known Activations