INDEX

Explanations

truth-related statements

oai_token-act-pair · gpt-3.5-turbo

references to truth and subjective opinions

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

obook

-0.70

 Mehran

-0.63

etsk

-0.63

riers

-0.63

oplan

-0.57

naires

-0.57

 Banner

-0.56

({

-0.56

enance

-0.55

 Skydragon

-0.53

POSITIVE LOGITS

ya

0.97

arently

0.96

 wise

0.95

ortunately

0.93

cially

0.90

identally

0.90

 caveat

0.90

wise

0.89

pecially

0.88

nown

0.86

Activations Density 0.214%

truth-related statements

references to truth and subjective opinions

No Comments

No Known Activations