INDEX

Explanations

references to high stakes, risks, or significant consequences in a situation

oai_token-act-pair · gpt-3.5-turbo

references to "stakes" in various contexts

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

issance

-0.84

ufact

-0.82

ICA

-0.82

vironment

-0.77

ergy

-0.75

lez

-0.74

uve

-0.72

eric

-0.72

ica

-0.71

audi

-0.68

POSITIVE LOGITS

 stakes

1.06

holder

1.03

rament

0.73

 escalation

0.70

cape

0.69

 ante

0.69

 stake

0.68

 roller

0.67

holders

0.66

pole

0.66

Activations Density 0.012%

references to high stakes, risks, or significant consequences in a situation

references to "stakes" in various contexts

No Comments

No Known Activations