INDEX

Explanations

names of individuals who have been sentenced or convicted of various crimes

oai_token-act-pair · gpt-3.5-turbo

instances of the verb "was" 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 @ 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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

inventoryQuantity

-0.76

PsyNetMessage

-0.74

pedia

-0.71

 congr

-0.69

 intrusion

-0.64

yip

-0.64

odes

-0.64

 certs

-0.63

 clicks

-0.63

 Dynamics

-0.63

POSITIVE LOGITS

 sentenced

1.67

 acquitted

1.60

 convicted

1.57

 indicted

1.56

 jailed

1.47

 charged

1.45

 arra

1.42

 extrad

1.37

 arrested

1.33

 accused

1.30

Activations Density 0.136%

names of individuals who have been sentenced or convicted of various crimes

instances of the verb "was" in various contexts

No Comments

No Known Activations