INDEX

Explanations

words related to authoritative control or influence

oai_token-act-pair · gpt-3.5-turbo

references to the concept of power and authority

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

Von

-0.79

TAG

-0.71

 Taste

-0.70

ead

-0.68

ALK

-0.68

uran

-0.67

riad

-0.66

eryl

-0.66

dk

-0.65

 Wheat

-0.65

POSITIVE LOGITS

 vested

1.00

 levers

0.99

houses

0.96

 outage

0.88

 wielded

0.84

 conferred

0.84

lifting

0.82

stroke

0.82

FUL

0.81

 exerc

0.79

Activations Density 0.042%

words related to authoritative control or influence

references to the concept of power and authority

No Comments

No Known Activations