INDEX

Explanations

references to logical reasoning and arguments

oai_token-act-pair · gpt-3.5-turbo

references to logic, particularly in legal and philosophical contexts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

semble

-0.70

orks

-0.70

 Volunte

-0.69

avez

-0.68

ometown

-0.68

hold

-0.65

eneg

-0.65

lain

-0.64

affer

-0.64

Shar

-0.63

POSITIVE LOGITS

 logic

1.31

 Logic

0.98

 droid

0.84

istically

0.83

 appl

0.81

matical

0.76

ynes

0.76

idi

0.72

 reasoning

0.71

matically

0.71

Activations Density 0.007%

references to logical reasoning and arguments

references to logic, particularly in legal and philosophical contexts

No Comments

No Known Activations