INDEX

Explanations

phrases mentioning contradictory information or beliefs

oai_token-act-pair · gpt-3.5-turbo

the word "contrary" and its variations, indicating a focus on presenting opposing viewpoints or rebutting popular beliefs

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

aquin

-0.78

urated

-0.78

ahime

-0.77

ilitating

-0.75

arnaev

-0.71

beans

-0.70

afer

-0.70

artney

-0.69

hens

-0.69

istan

-0.69

POSITIVE LOGITS

etheless

0.91

 minded

0.85

lihood

0.79

ly

0.74

 notwithstanding

0.72

 chronological

0.72

 contrary

0.71

lly

0.71

 guiActiveUn

0.71

ptions

0.70

Activations Density 0.012%

phrases mentioning contradictory information or beliefs

the word "contrary" and its variations, indicating a focus on presenting opposing viewpoints or rebutting popular beliefs

No Comments

No Known Activations