INDEX

Explanations

references to legal actions and court sentences

oai_token-act-pair · gpt-3.5-turbo

instances of the word "sentenced" in relation to legal consequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

RIS

-0.78

walker

-0.76

aucus

-0.75

Champ

-0.68

opic

-0.67

WHERE

-0.66

ÃŁ

-0.64

upp

-0.62

ween

-0.61

 Piercing

-0.61

POSITIVE LOGITS

 sentenced

1.27

icts

0.99

 sentencing

0.97

 convict

0.93

 inmates

0.89

 confir

0.86

 felon

0.85

 punished

0.84

 convicted

0.84

 punishable

0.84

Activations Density 0.009%

references to legal actions and court sentences

instances of the word "sentenced" in relation to legal consequences

No Comments

No Known Activations